research paper about face to face classes

Cornell Chronicle

• Architecture & Design
• Arts & Humanities
• Business, Economics & Entrepreneurship
• Computing & Information Sciences
• Energy, Environment & Sustainability
• Food & Agriculture
• Global Reach
• Health, Nutrition & Medicine
• Law, Government & Public Policy
• Life Sciences & Veterinary Medicine
• Physical Sciences & Engineering
• Social & Behavioral Sciences
• Coronavirus
• News & Events
• Public Engagement
• New York City
• Photos of the Week
• Big Red Sports
• Freedom of Expression
• Student Life
• University Statements
• Around Cornell
• All Stories
• In the News
• Expert Quotes
• Cornellians

Face-to-face interaction enhances learning, innovation

By james dean.

Whether virtual or in-person learning is better may be the wrong question.

New Cornell psychology research finds that sitting face-to-face, rather than shoulder-to-shoulder, enhances learning and innovation – even when we’re learning complex physical skills that should be harder from that perspective.

In experiments, children and adults solved a complex visual and spatial problem – opening a puzzle box – more quickly after they watched a model demonstrate a solution face-to-face, compared with others who observed from next to or perpendicular to the model.

Across ages, test subjects performed better when they could observe not only an instructor’s hands but also their eyes, gaze and facial movements. The researchers propose that face-to-face interaction transmits valuable social information about goals and motivations in addition to visual information about the task.

“This shared mental perspective might be more important for some types of learning than sharing a common visual perspective,” said Ashley Ransom , M.A. ’18, a doctoral student in the field of developmental psychology. “Face-to-face interaction might facilitate creativity and innovation rather than strict mimicry of the teacher.”

Ransom is the lead author with Dr. Brian LaGrant ’17, M.D. ’21, of “ Face-to-Face Learning Enhances the Social Transmission of Information ,” published Feb. 25 in PLOS ONE. Adam K. Anderson , professor in the Department of Psychology and in the College of Human Ecology (CHE), and Eve De Rosa , associate professor of psychology and a Rebecca Q. and James C. Morgan Sesquicentennial Fellow in CHE, and the dean of faculty, are senior authors. Anthony Spiteri, a former postdoctoral fellow in Anderson and De Rosa’s Affect and Cognition Lab , and Tamar Kushnir, now a professor at Duke University, are co-authors.

Learning a new visuospatial task, such as how to tie a knot or play an instrument, is thought to require us to adopt the teachers’ perspectives, to try to see the world through their eyes. However, the new research suggests it might also be important to actually see their eyes.

Past studies have shown that imitating specific steps is easier when learners see what an instructor sees – for example, watching a video of hands building a circuit board. As we move from that 0-degree viewpoint to an opposite angle at 180 degrees, our brains must employ “mental rotation” to understand a model’s movements as meaning left or right, forward or backward.

As a result, the researchers said, learning to open a puzzle box should be more difficult when observing someone face-to-face. But the study found otherwise.

“Remarkably, the simple act of sitting across from someone can help overcome limitations in shared visual perspective,” the authors wrote. “Face-to-face learning overrode the inherent difficulty of taking another’s visual perspective.”

The colorful puzzle box – also called an artificial fruit box when first developed for studies with chimpanzees – contained four layers requiring 12 steps to open, including removing bolts and panels, moving sliders and turning screws. There were multiple possible solutions, and the model incorporated superfluous movements in their demonstration to help measure imitation.

The researchers randomly assigned 36 children ages 4 to 6, and 57 university students ages 18 to 27, to watch demonstrations from 0-, 90- or 180-degree orientations relative to the model. The subjects then got three tries to open the box from whatever position they preferred.

By one standard, face-to-face learners were less successful: They imitated less faithfully than participants who couldn’t directly see faces. But in accomplishing the real goal of opening the puzzle box, they were faster and more likely to come up with new solutions rather than relying on the model’s solution.

“They weren’t as good at mimicking, but there’s a benefit to that because it facilitated discovery,” Anderson said. “A social perspective – looking at people and where they look – allowed children and adults to become better learners at the condition that should have been the most challenging.”

Adults overall were more likely to copy the model’s solution and to choose to solve the puzzle box from the teacher’s vantage point. Children, in contrast, came up with new solutions and often remained in their original position.

Those results suggested adults had become better rote learners but less innovative with time and more formal education.

“Adults focused on recreating the model’s actions rather than the end result,” the authors wrote. “Children are more flexible learners than adults and engage in more exploration during learning.”

The research was supported by a grant awarded to De Rosa from the State University of New York’s Empire Innovation Program.

Media Contact

Gillian smith.

Get Cornell news delivered right to your inbox.

You might also like

Gallery Heading

ORIGINAL RESEARCH article

Face-to-face or face-to-screen undergraduates' opinions and test performance in classroom vs. online learning.

• Discipline of Psychology, School of Medicine, University of Tasmania, Hobart, TAS, Australia

As electronic communication becomes increasingly common, and as students juggle study, work, and family life, many universities are offering their students more flexible learning opportunities. Classes once delivered face-to-face are often replaced by online activities and discussions. However, there is little research comparing students' experience and learning in these two modalities. The aim of this study was to compare undergraduates' preference for, and academic performance on, class material and assessment presented online vs. in traditional classrooms. Psychology students ( N = 67) at an Australian university completed written exercises, a class discussion, and a written test on two academic topics. The activities for one topic were conducted face-to-face, and the other online, with topics counterbalanced across two groups. The results showed that students preferred to complete activities face-to-face rather than online, but there was no significant difference in their test performance in the two modalities. In their written responses, students expressed a strong preference for class discussions to be conducted face-to-face, reporting that they felt more engaged, and received more immediate feedback, than in online discussion. A follow-up study with a separate group ( N = 37) confirmed that although students appreciated the convenience of completing written activities online in their own time, they also strongly preferred to discuss course content with peers in the classroom rather than online. It is concluded that online and face-to-face activities can lead to similar levels of academic performance, but that students would rather do written activities online but engage in discussion in person. Course developers could aim to structure classes so that students can benefit from both the flexibility of online learning, and the greater engagement experienced in face-to-face discussion.

Introduction

With changing student lifestyles and fast-developing technology, universities are increasingly offering more “flexible” learning environments. Commensurate with the opportunities that technological advances afford, for over a decade ( Imel, 2002 ) the provision of online, e-learning experiences has undergone rapid expansion in the higher education sector. Today, online learning is part of the student experience for a substantial proportion of university students in a variety of countries (e.g., Ituma, 2011 ; Otter et al., 2013 ; Tucker et al., 2013 ). The current study aimed to compare students' experience and performance in both online and traditional face-to-face learning experiences.

The rise of e-learning has helped to encourage students to take on more responsibility for their own acquisition of knowledge ( Ituma, 2011 ). In a traditional, teacher-centered model of teaching, the lecturer transmits knowledge to students, with little input from those students ( Harden and Crosby, 2000 ; Prosser et al., 2005 ). However, the shift to less traditional classes has coincided with a greater focus on more student-centered learning, with the lecturer facilitating or managing the students' learning, rather than simply transmitting information ( Balluerka et al., 2008 ). Because of the more self-directed learning assumed to occur in online environments, online learning may have the potential to produce more in-depth discussions and to improve the quality of learning, as well as having the practical benefits of encouraging wider student participation and increasing the cost-effectiveness of education, compared to traditional face-to-face learning ( Smith and Hardaker, 2000 ; Alexander, 2001 ). A timely example is that of flipped classrooms, whereby the students engage in active learning (often via vodcasts or in online discussions) and the instructor provides support and scaffolding ( Strayer, 2012 ).

Given the potential pedagogical advantages of online learning, there is a pressing need to formulate an evidence-based understanding of best practice in this area. However, despite the increasing research interest in e-learning, there seems to be little consistency in the training that lecturers receive in developing online materials. This may be an artifact of a diverse research base. For example, many studies have focused on the efficiency, content and delivery method that teaching staff have developed (e.g., Rossman, 1999 ; Twigg, 2003 ; O'Neill et al., 2004 ), while the perceptions and experiences of the students themselves have been largely neglected ( Alexander, 2001 ; Holley and Oliver, 2010 ; Ituma, 2011 ). Some teaching staff seem to perceive web-based platforms simply as an alternative method for presenting the traditional content, whereas others might look for more innovative ways of using such platforms to improve students' engagement and thus their learning outcomes ( Holley and Oliver, 2010 ; Ituma, 2011 ). Much of the initial experience of e-learning failed to live up to learners' expectations ( Imel, 2002 ), partly because of technological constraints, and partly because of the early instructional approaches taken ( Anderson and Dron, 2011 ). Even today, students in different courses can have quite different experiences of online learning. There is a clear need for more research into what does and does not work in online learning, but also for a focus on the student experience in the increasingly digital landscape of tertiary education.

The worldwide trend toward online learning provision has resulted in numerous online-only courses, and universities in Australia are no exception (e.g., Bell et al., 2002 ; Tucker et al., 2013 ). However, the student experience in online classes is a different one from in traditional face-to-face classes, and patterns of engagement seem to differ between the two ( Robinson and Hullinger, 2008 ). For example, Otter et al. (2013) found that students in online-only classes felt more disconnected from their peers and lecturers, more obliged to be self-directed in their studies, and less aided by their lecturer, than their lecturers believe them to be. Students can also feel daunted by the technological expectations of online study, especially if they start off without sufficient technical knowledge or support ( Zhang and Perris, 2004 ; Holley and Oliver, 2010 ).

Most of the research to date has focused on courses offered entirely online. Yet, an increasing number of face-to-face courses are beginning to incorporate some online components, in which traditional in-class activities are supplemented, rather than replaced, with online activities ( Ituma, 2011 ). However, very little is known how students respond to this kind of “blended e-learning,” especially when they have enrolled in a face-to-face class, rather than one which they expect in advance to be completed online. This represents a substantial gap in the educational literature, as there are potentially important implications for student engagement, performance, and attrition. In general, student engagement in traditional classes is positively associated with student engagement and academic performance, although the magnitude of these effects may be small (e.g., Carini et al., 2006 ). Some research suggests that participation in learning technology can itself increase engagement and learning ( Chen et al., 2010 ), and flipped classrooms are emerging as a promising student-centered paradigm (e.g., Galway et al., 2014 ). However, a deeper consideration of these positive outcomes is warranted, as outcomes seem to depend on the nature of the online activities, and the way that students interact with them. For example, Davies and Graff (2005) found that students who interacted and participated more in online discussion did not show significantly better academic performance than students who were less involved in that discussion. In contrast, Evans et al. (2004) showed that students performed much better when their online course material was accessible via an interactive, navigable format than via a series of scrollable web-pages. Thus, there is mixed evidence about the extent to which some online activities might help or hinder students' marks, compared to more traditional, face-to-face classes.

There are various reasons to expect that students might prefer at least some aspects of online learning to traditional classes. Previous researchers have suggested that in contrast to the faster, real-time pace of face-to-face classes, the extra time available for online activities might allow students to think about course material more critically and reflectively, leading to deeper understanding of the course content (e.g., Ramsden, 1992 ; Robinson and Hullinger, 2008 ). Others have suggested that the less confrontational or personal nature of e-learning might encourage shyer students to engage more, or to feel less pressure than in face-to-face interactions ( Warschauer, 1997 ; Hobbs, 2002 ).

However, there are also reasons for which students might prefer more traditional, in-class activities. Although social connectedness can be derived online ( Grieve et al., 2013 ), most students feel that face-to-face contact is essential for building a sense of community ( Conole et al., 2008 ). Even when classes are only partially online, students may feel that online discussion detracts from this feeling of community with their peers and tutor. Further, at a practical level, students need to exercise more self-motivation to complete activities online, compared to in-class, where that role of motivator is taken on by the lecturer ( Upton, 2006 ). Thus, it is important to explore students' perceptions of both online and face-to-face learning experiences, rather than just one or the other.

At our university, there is currently a focus on incorporating more online material into face-to-face units, by replacing some weekly practical classes with self-directed, online activities. The primary motivation is a financial one, as online classes do not require paid tutors, and lecturing staff are not paid any more for developing online material than in-class material. A secondary motivation is the general assumption that students prefer the flexibility and opportunity for self-directed learning provided by online activities. However, there does not seem to have been any assessment of students' academic outcomes, or their overall experience, in terms of online learning within this context.

The current research was therefore designed to examine the performance and perceptions, in both face-to-face and online learning experiences, of Australian undergraduate students who were enrolled in traditional face-to-face units which incorporated some online components. We took a two-fold approach, with the aim of providing an integrated picture of both objective and subjective outcomes. To this end, we compared both students' academic performance and their qualitative comments on their learning experience, between offline and online modalities, as described below.

In order to control for individual differences and thus increase the statistical power of the study, we tested the same students on both offline and online tasks, rather than having separate groups for each modality. Alongside the empirical rigor of this within-groups design, we wished to maintain the authenticity of the measures, and thus chose to embed the tasks within the students' actual class learning experience, rather than in an artificial, lab-based study. We also wanted to ensure that instead of just observing students' behavior (e.g., counting their discussion contributions), we assessed both objective and subjective measures of their learning. The specific research questions were thus as follows:

After having engaged with course content either face-to-face or online, would students differ in:

1. Their academic marks on a short test of the concepts just learned?

2. Their preference ratings for online and face-to-face modalities?

3. The themes emerging in their justifications of these ratings?

In the first study, undergraduate psychology studied two carefully matched topics, one in a face-to-face class and the other online, with the order counterbalanced between two groups. We compared marks on the two topics, and examined the students' comments on their experience of completing the activities, one in-class and one online.

Participants

Participants were 67 third-year (advanced level) undergraduate students at an Australian university, 13 male and 54 female. Their mean age was 24 years ( SD 7.1 years), and all spoke English as a first language. The students took part in the study as part of their unit on developmental psychology, but gave consent for their data also to be used for research purposes. All were familiar and confident with the web-based platform used for delivering the online activities. The study received approval from the university's Human Research Ethics Committee.

Materials and Procedure

Participants completed written exercises, a class discussion, and a written test (worth 5% of their overall mark) on two academic topics, one week apart. The first topic concerned children's stages of cognitive development, and the second, children's stages of drawing development. Three classes of students participated. The classes were allocated to two groups (as described in the next paragraph), with the larger class allocated to Group 1 ( n = 27), and the two smaller classes to Group 2 ( n = 40). This meant that the two groups differed in size. However, there were no significant differences between the two groups in terms of age [ M = 23.2 years ( SD 6.2) and M = 26.0 years ( SD 8.2), respectively], sex distribution (80 and 82% female), or overall performance in this unit [ M = 62.4% ( SD 11.6%) and M = 65.0% ( SD 10.0%)], all p s > 0.1.

In Week 1, both groups of participants completed the face-to-face activities in their weekly practical class. Group 1 students discussed their recent experience of observing a child completing several tasks of conservation of matter (e.g., conservation of liquid, conservation of number, Piaget, 1954 ). The previous week, each student had individually administered these tasks to a 5- to 12-year-old child of their acquaintance. The few students who did not know a child to visit had viewed multiple online videos of children being administered these tasks and chosen the one on which they would like to answer the test questions. There were no differences in performance between these two sets of students. Group 2 students were given a series of self-portraits drawn by children aged 2–16 years, and a copy of Lowenfeld's (1939) stages of drawing development.

In class, the tutor initiated the discussion with more general questions, for example (for conservation, Group 1): “What did you find most difficult about giving the task, and how did you overcome this difficulty?,” and (for drawing, Group 2) “Do you think it would be easy or difficult to persuade a child of this age to produce a self-portrait? Do you think it would be easier with older or younger children?.” The questions later narrowed in on theory, with the tutor asking the students in Group 1 to discuss which cognitive stage they thought the child was in (in terms of Piaget's stage model of cognitive development, e.g., pre-operations or concrete operations), and how the wording of the adult's instructions could influence the child's performance. In Group 2, the tutor asked the students to decide which developmental stage each self-portrait seemed to represent (e.g., pre-schematic drawing, schematic drawing), and why. They also talked about the motor skills required for drawing. For both groups, discussions took place first in small groups and then as a whole class, over a 1-h period, with the questions provided by the tutor but the discussion structured by the students themselves. At the end of the class, both groups were then given a half-hour, in-class, written test on the topic that they had just studied (i.e., development of conservation or drawing):

1. Participants answered six short-answer questions structured as similarly as possible for the two topics (e.g., Conservation : Based on your observation, which Piagetian stage do you think this child is in, and why? Drawing: Based on the drawing, which of Lowenfeld's stages do you think this child is in, and why?)

2. Participants rated their preference for completing these activities face-to-face vs. online, on a five-point scale ranging from 1 (“much prefer to do in class”) to 5 (“much prefer to do online”), and were asked to identify “one thing that was good about doing the activities in class, not online,” and “one thing that would have been good about doing the activities online, not in class.”

During the following week (Week 2), both groups of participants completed the online activities at a time of their own choosing, for the topic that they had not covered in the previous week. Group 1 scrolled through the series of children's self-portraits, presented in PDF form, and were provided with an online copy of Lowenfeld's (1939) stages of drawing development. Group 2 were asked to think about the conservation tasks they had observed.

Both groups were asked to consider the same questions that had been considered in the face-to-face classes as described above, for drawing development (Group 1) and for cognitive development (Group 2). Students were then instructed to go to the online discussion board. They did not have to transcribe their answers for all of the questions that they had considered, but were asked to contribute their answers to the questions that they “had found most interesting,” or to explain any questions that they “had found particularly difficult, and why.” Students were also asked to read others' contributions, and to respond to them, “linking their experiences and answers with yours if you can.”

The opportunity to reflect on the materials and questions, and to contribute to the online discussion, was available for 5 days, with each participant contributing at least once. Each participant contributed to the discussion, as required (with at least one discussion post that covered several questions). The day after the online discussion board was closed, the online version of the test was made available, with the same questions as in the face-to-face version:

1. Participants responded in writing to the same six short-answer questions answered in class the week before.

2. Participants rated their preference for completing these activities face-to-face or online, and were asked to identify “one thing that was good about doing the activities online, not in class,” and “one thing that would have been good about doing the activities in class, not online.”

Students had 4 days in which to do the test, which was open for 30 min once it was begun. It is possible that the students completing the test online could have accessed extra information (from the internet) than the students completing the test in class (without internet access). However, trying to find further information during the limited time available for writing the test answers would not have been very helpful. The questions were focused on discussing the conservation or drawing performance of “their” child, and all students (in class and online) had access to a copy of Lowenfeld's stages, and all had been aware that they would need to revise the basics of Piaget's cognitive stage theory for the test.

In summary, both groups completed the same activities, discussion and tests, but in Week 1 both groups did so in face-to-face classes (on cognitive development for Group 1 and on drawing development for Group 2) and in Week 2 both groups did so online (on drawing development for Group 2 and on cognitive development for Group 1).

Results and Discussion

Class marks and preferences.

Participants' marks on the two tests, as well as their stated preferences for studying each topic face-to-face or online, were calculated. Table 1 shows the means and standard deviations on each of these factors, for participants in the two groups. As seen in the table, students showed an overall preference for studying topics face-to-face rather than online, with their average preferences of around 2 hovering much closer to the “much prefer to do in class” end of the 5-point scale than the “much prefer to do online” end of the scale.

Table 1. Study 1: Mean Marks and Modality Preferences across Task and Modality .

A series of repeated-measures analysis of variance (ANOVAs) showed that there were no significant differences between the groups on any of the comparisons made. The two groups did not differ significantly in the marks they gained, in terms of either the topic they were tested on [ F (1, 65) = 0.67, p = 0.42, partial η 2 = 0.01] or modality in which this test was presented [ F (1, 65) = 0.31, p = 0.58, partial η 2 = 0.01]. Nor did they differ significantly in their modality preference, in terms of either the topic about which they were asked [ F (1, 65) = 0.42, p = 0.52, partial η 2 = 0.01] or the modality in which they were asked it [ F (1, 65) = 0.05, p = 0.82, partial η 2 = 0.001]. We followed up this significant group difference in modality preference by also considering whether there was a significant difference in terms of the number of students who (much) preferred to do the tasks in class and those who (much) preferred to do them online/did not mind either way. A chi-squared analysis confirmed that significantly more students preferred to complete the tasks in class ( n = 47) than online/didn't mind ( n = 20), χ 2 (1) = 10.89, p < 0.001.

Pearson correlations were calculated to see if there was any association between participants' academic performance and their preference for learning face-to-face vs. online. We first considered participants' scores on the two topic-specific tests and their reported modality preference for each topic. The correlations were not significant for either the topic done online ( r =−0.02, p = 0.85), or the topic done in class ( r = 0.01, p = 0.95). This suggests that students did not simply prefer a modality because they felt they had performed better on the task in that modality, and nor did they perform better in the modality that they already preferred. We then looked at participants' final mark on this entire academic unit ( M = 64.0%, SD = 10.7%), and calculated the correlation between this mark and their modality preference averaged over the face-to-face and online tasks ( M = 2.09, SD = 0.92). The correlation was not significant ( r = −0.11, p = 0.40), suggesting no consistent relationship between overall academic performance and preference for learning online or in class. Finally, we confirmed that there was no significant correlation between participant age and modality preference ( r = 0.01, p = 0.97), nor between age and overall marks ( r = 0.05, p = 0.69).

Qualitative comments

We also asked participants to note one thing that they liked better about completing the activities (exercises, discussion, and test) face-to-face, and one thing about what they liked better about completing them online. Participants provided their own freeform answers, which we then subjected to thematic analysis ( Braun and Clarke, 2006 ). We chose thematic analysis rather than a more specific approach such as Interpretive Phenomenological Analysis (IPA) or Discourse Analysis, as we wished to identify repeated patterns of meaning, or themes, in the responses, but without attaching our analysis to any pre-existing theoretical framework. We followed Braun and Clarke's (2006) six phases to identify the themes as they emerged from an analysis of students' responses, rather than pre-existing themes being imposed on a grouping of the responses. Specifically, we (1) familiarized ourselves with the scope and nature of the students' responses to both the “in-class” and “online” questions, and (2) generated initial codes on the basis of a subset of 25 participants' responses, by identifying the overall point made in each response with a 2- to 4-word summary code, which we (3) collated into tentative themes and modified as we reviewed another 25 responses. These potential themes were then (4) reviewed to check how they fitted with the entire data set, and (5) the themes were refined into an exhaustive set of final themes for all responses, which were named and clearly defined. Finally, (6) the most representative example for each theme was selected, and the final analysis related back to the research questions. Although participants had been asked to provide only one reason for liking something about each modality, many provided two or even more reasons. We included the first one or two reasons given in our thematic analysis. Table 2 shows the proportion of responses that were categorized into each them, for face-to-face and online learning, for the whole sample, n = 67.

Table 2. Study 1: Proportion of Responses per Theme, for Face-to-Face and Online Learning .

In justifying their modality preferences about completing the activities face-to-face, the 67 participants provided 102 reasons overall. The themes that emerged were as follows:

• More engagement: This was the theme with the most responses, with students noting that they felt more engaged when the activities were completed in the social environment of a classroom setting, rather than online. Most also commented that they felt this greater engagement led to more meaningful discussion, for example, I think that discussion face to face really allows you to think more deeply and bounce ideas of other people. Writing it online, felt like your answers had to be more formal and exact, whereas in class discussion I felt you could really bounce more possible ideas off each other before coming to a conclusion.

• Immediate feedback: The next most common theme was that participants appreciated the fact that each comment they made in class immediately elicited a related comment from a peer, or a clarification from the tutor in real time, rather than having to wait hours for a response to their particular comment online. An illustrative example is, You are able to directly discuss with tutor and peers and therefore directly receive feedback for your questions and others questions.

Overall, 80% of the first reasons (and 37% of second reasons) that participants gave to illustrate the benefits of in-class activities fitted these themes of “more engagement” and “immediate feedback.” Some of remaining responses referred to more practical considerations:

• No wish to read comments: Some students noted that they had no wish to read the comments of their classmates online, although they were happy to listen to and interact with their classmates in real life.

• Easier to review paper documents: Others noted it was easier to review material in class, because they could spread out the pages in front of them on a table rather than scrolling through on screen. The other remaining responses fitted the more pedagogically motivated themes that in-class interaction provided:

• Deeper understanding and a

• Better flow of argument .

When explaining their preferred modality for online activities, the 67 participants provided 78 reasons, rather fewer than the 102 that they had given about face-to-face classes.

• Convenience : The most common theme was a practical one; the greater convenience of being able to complete the online activities in their own time (within the given week), and/or at any location, as in the comment, Can do it at anytime, and anywhere.

• Wider contributions: The second most common theme was that the online discussion allowed contributions from a greater range of people than a class discussion, in which shyer students sometimes stayed quiet in the presence of their more outspoken peers, for example, Discussion not dominated by loud, confident people.

• More detailed responses: Students also noted that the online discussion forum encouraged more detailed responses than in-class discussion (where conversational turns were typically shorter, but more frequent, than in writing), as in the response, I thought it was good to be able to read everyone's experiences with full details, as time restraints in class don't allow for each individual to thoroughly go through their task.

Together, 84% of first reasons (and 69% of second reasons) fitted these three themes. The remaining responses were categorized into smaller themes; specifically:

• More time to think : Some students noted that providing responses online gave them more time to consider their answers than having to speak spontaneously in class.

• Faster to type : It was noted that it was faster to provide answers to the test when it was typed online than handwritten in the classroom.

• Less judgment : Finally, some students felt less judged by their tutor and peers when answers could be written than spoken.

Overall, many of the reasons that students gave for each modality differed from those seen in previous literature, as is taken up again in the General Discussion.

In Study 1, students had been asked to reflect on their preferences and perceptions of several different activities simultaneously: written exercises, class discussion, and a short test. However, the study was not set up to allow students to rate whether they would prefer to do some of these activities in one modality and others in another (these preferences emerged only in the qualitative comments). Further, participants also had to come up with “one good thing” about both face-to-face and online learning, even though some responses suggested that students had simply felt obliged to come up with benefits that they did not really see as an advantage, for example, [Online activity] can be done in the students own time… but since there is a dedicated time each week for a practical class, this kind of negates this advantage .

In order to assess students' ratings of the activities separately, and to ask them only for the reasons why they preferred one modality to another, we conducted a second study. In this follow-up, we asked a new group of undergraduates to reflect on their preferences for doing two separate activity types face-to-face or online: short written exercises to get students thinking about the topic (which would traditionally have been done in class but could just as well be done alone, in students' own time), and class discussions.

Participants were 37 third-year undergraduate psychology students (10 male) at the same Australian university, who had not participated in Study 1. Their mean age was 25 years ( SD 8.9 years). All but one spoke English as their first language, and they participated as part of their course requirements. As in Study 1, all were competent and familiar with the technical requirements of the online learning activities. Ethics approval was provided by the university's Human Research Ethics Committee, and participants gave consent for their data to be included in the research.

Participants were completing an undergraduate unit which included both face-to-face and online practical classes, and so they were accustomed to both modalities. In this particular unit (Psychology of Language), students had participated in each type of practical class in the last 2 weeks. In both the first week (class conducted face-to-face) and the second week (class conducted online), students had completed written activities, and participated in a discussion, and so the experience of both should have been relatively fresh in their minds. Specifically, in the second week's online practical class, students had completed written activities on language and gender (writing down the best word to address unknown people of varying age/sex, e.g., sir , mate , miss ), language in the digital age (responding to fictitious text messages), and language as a marketing tool (comparing two university courses on the basis of the universities' self-descriptions). That same week, students then contributed to an online discussion about their responses to these tasks. The following week, in a face-to-face class, participants discussed their responses to the previous week's tasks. They were then asked to take a few minutes to consider the advantages and disadvantages of completing the written activities online, rather than in a face-to-face class. Even though the students had not done these particular written activities in class, only online, we considered that they would be easily able to make this judgment, because they had previous experience with doing similar kinds of written activities in class on multiple occasions. Students were also asked to consider the advantages and disadvantages of engaging in discussion of these particular language-based issues in class, and online (both of which they had experienced).

Finally, students were asked to:

1. Rate, on a scale of 1 (“much prefer to do in class”) to 5 (“much prefer to do online”), how they would prefer to complete these written activities , and to write down their main reason for this rating.

2. Rate, on the same scale, how they would prefer to engage in this class discussion , and to write down their meaning reason for this rating.

Preferences

Overall, participants showed a mean preference for written activities that was close to the middle of the 5-point scale, 3.54 ( SD = 1.37). In contrast, their mean preference for discussions was closer to the “much prefer to do in class” end of the scale, 1.62 ( SD = 1.01). The difference between the two ratings was significant, F (1, 36) = 59.3, p < 0.001. It might be expected that some students simply preferred to do all activities in class or online, but this did not appear to be the case: there was no significant correlation between modality preferences for written activities and discussion, r = 0.21, p = 0.21. Nor was there any significant correlation with age and either of the modality preferences, written activities r = 0.17, p = 0.31 or discussion, r = 0.06, p = 0.71. Thus, it was clear that students had different preferences according to the nature of the task, preferring to do written activities online but discussions face-to-face.

We also asked participants to note the main reason for their ratings. Using the same process as in Study 1, we conducted a thematic analysis on the responses. Again, some participants provided two reasons, and we coded both of these.

Table 3 shows the themes which came out of the responses about preferences for written activities. As shown in the top panel, only 9 participants (much) preferred doing the written activities in class, while 21 (much) preferred doing the written activities online, as shown in the bottom panel. (The responses from the 5 participants who didn't mind either way were omitted.) As in Study 1, we conducted a chi-square analysis, and confirmed that significantly more participants preferred to do written activities online (or did not mind either way) than to do them in class, χ 2 (1) = 8.26, p < 0.005.

Table 3. Study 2: Proportion of Responses per Theme for Written Activities, for those who Preferred these Face-to-Face vs. Online .

The themes emerging from the responses of the relatively small group of participants who preferred doing written activities in class were:

• Interaction encourages learning : The students noted that the interaction of face-to-face classes encouraged learning, in a way that could not be achieved via the solitary completion of the activities, for example, Because if feels like more active learning in class, interacting and talking with others.

• Class time already allocated : A second theme was that since the class time was already allocated for doing activities, students would rather do the work in class, with others, rather than have to force themselves to do it in their own time, as in, Because I find it hard to make time for non-attending pracs and when I do I rush through them as fast as I can. Whereas if I have to come to class I'm more likely to concentrate and understand what I'm doing.

Finally, some students valued the

• Immediate feedback offered in response to any comments they made

• Deeper learning that occurred in class, where they felt more focused on the activities.

The larger group of participants who preferred doing written activities online overwhelmingly reported the following theme:

• Convenience : Students noted the convenience of being able to complete the work at a time and place that suited them, as in, If its online I can do it in bed .

Other responses fitted the themes of finding that doing the activities online provided:

• More time to think

• Less distraction

Finally, some students noted that it depended on the nature of the written activities—but when they were reasonably challenging but not really difficult (as the current activities were perceived), they were the

• Right level for doing online, and

• Would waste time if done in class as individual work.

The themes which emerged from participants' preferences about class discussions are summarized in Table 4 . Almost all participants (much) preferred discussions in class ( n = 26, top panel), with very few (much) preferring them online ( n = 3). (The 6 participants who didn't mind either way were not included in this analysis.) A chi-square analysis confirmed that significantly more students preferred to engage in discussions in the classroom than online (or did not mind either way), χ 2 (1) = 8.26, p < 0.005. There was a range of reasons for preferring in-class discussion, with the following themes:

• More engagement: The most common theme to emerge was that it allowed more engagement than online discussion, for example, I would rather be in a classroom talking to actual people and engaging more.

• Better flow of discussion: Another popular theme was that discussion flowed better in person than online, as in, Can actually have a free flowing conversation .

• Personal setting: Some participants noted that the personal setting of the classroom encouraged better discussion than the more impersonal online environment, for example, The discussion in person is more beneficial for learning. Easier to communicate and express ideas in a personal setting .

• Greater range of opinions: Students expressed that in-class discussion exposed them to more opinions, for example, Get more opinions and discussion of them in class .

Table 4. Study 2: Proportion of Responses per Theme for Discussion, for those who Preferred Discussion Face-to-Face vs. Online .

The remaining comments expressed the themes of:

• No wish to read others' comments (although they were happy to listen to them)

• Immediate feedback of face-to-face discussion

• Deeper learning facilitated by in-class discussion

• Time already allocated in class, so students were happy to do the discussion then.

As shown in Table 4 , only three participants said that they preferred discussions online, each with a separate reason.

General Discussion

The goal of this research was to explore whether students' academic marks and their preferences for learning varied as a function of online or offline delivery. A further aim was to illuminate the “active ingredients” contributing to students' experiences in these modalities, using a qualitative methodology. In two studies, undergraduate students engaged in practical class activities in two modalities: in a face-to-face class, and online. We assessed students' academic performance on tests of class content and asked students which modality they preferred for learning, and why. Overall, we saw no clear differences in academic performance in online vs. in-class learning, but students had a general preference for in-class activities, specifically when discussion of academic topics was required.

In terms of academic performance, our first study showed that there were no significant differences in test performance whether class material, and the subsequent test, was presented face-to-face or online. This finding adds to previous evidence that simply participating in online activities does not necessarily lead to significantly improved test scores ( Davies and Graff, 2005 ). On the surface, these findings are difficult to explain using student-centered models of e-learning: it is assumed that the self-directed learning that occurs in in online environments should result in deeper learning. However, an alternative explanation comes from Garrison (2012) , who notes that the community of inquiry created by e-learning is not synonymous with the collaborative and constructivist approaches that can foster deep learning. If so, perhaps the current findings reflect collaboration and constructivism in both modalities, but in regards to different aspects of the learner experience. Overall, performance across the two modalities is convergent, but the electronic and face-to-face pathways to that performance may be divergent. For example, the benefits obtained a student who has “more time to think” when working asynchronously online may be similar to the benefits obtained by another student “exposed to more opinions” within a classroom discussion. Future research could aim to unpack such benefits and their relationship with student characteristics: this could potentially results in truly student-centered approaches through the creation of bespoke learning activities.

There was also no significant link between the academic mark achieved on the unit overall and preference for either face-to-face or online learning activities. The lack of relationship between preference and performance is a notable one. At least in the current study, it seems that a liking for a particular modality does not benefit performance in that modality. Importantly, these findings also suggest that asking students to engage in their non-preferred modality does not mean that poor performance will ensue. The implication is that educators can select delivery modality based solely on pedagogical reasons with confidence, rather than being concerned that a certain delivery method may disadvantage students who hold negative perceptions of that method.

The participants in our first study preferred, on average, to complete activities in class rather than online. In the second study, participants had the opportunity to be more specific about their preferences. Their responses indicated that although they were happy to complete individual written exercises online (especially if these were of a reasonable level of difficulty), students much preferred to participate in discussions in face-to-face classes. Thematic analysis ( Braun and Clarke, 2006 ) of participants' qualitative comments in both Studies 1 and 2 revealed that students most liked the greater sense of engagement provided by face-to-face vs. online activities. Since there appears to be no previous direct comparison of students' perceptions of learning in these two modalities, it is difficult to link these responses with those in the literature concerning only online learning. However, it is interesting to note that the most common theme emerging in both studies was about the stronger feeling of engagement in the subject matter that these activities provided, with many students noting the deeper learning to come from this greater engagement. Although this result in no way negates previous findings that e-learning can also encourage deep learning (e.g., Ramsden, 1992 ; Robinson and Hullinger, 2008 ), it should be borne in mind that the benefits of e-learning do not diminish any pre-existing benefits of traditional classroom learning.

In Study 1, the immediate feedback from peers and tutor was also noted as an important advantage of in-class activities in general. In contrast, the immediacy of feedback available from online activities can vary greatly, depending on the nature of the activities and the way a unit is run. In the current study, students had 5 days in which to conduct their on-line discussion with peers. Thus, some students had to wait hours or even days for a peer to comment on their particular contribution, a situation that could not arise in an in-class discussion. In other units or courses, online discussions could be organized to occur at a set time, or as a “chat” function, in real time. This would allow contributions to be responded to more rapidly, and would presumably alleviate some of the frustrations that our participants experienced with more drawn-out discussions. In Study 2, in addition to preferring in-class discussion because of the greater engagement it offered, students also expressed the themes that face-to-face discussion allowed better flow of conversation, provided a more personal setting in which to enjoy the discussion, and encouraged a greater range of opinions from the group. These findings highlight the importance of facilitating collaboration and constructivism in learning.

When students were asked to note “one good thing” about online learning (Study 1), or to explain why they preferred to do written exercises online (the majority of participants in Study 2), thematic analysis revealed an important common theme: convenience. Students in both studies noted for online learning; they could do the work at a time of their own choosing, without having to travel to the university. This pragmatic reason emphasizes the advantages of online activities for many students, but does not fit with the more pedagogically driven reasons seen in previous research (e.g., Galway et al., 2014 ). Nevertheless, it does suggest that at least some students appreciate the flexibility in time and location afforded by online activities, especially as part of an otherwise face-to-face unit.

A number of participants in the first study also noted that online work encouraged contributions from a wider range of students, specifically, those who might be shy about face-to-face interaction. This reason has been noted in previous work (e.g., Citera, 1998 ; Hobbs, 2002 ). Nevertheless, on its own, this point does not provide a strong justification to adopt online learning, as many universities now consider oral communication skills as an important graduate attribute (e.g., Australian Qualifications Framework, 2013 ). Only a few students provided an academically oriented benefit of online learning, noting that it encouraged more detailed answers, and that contributors had more time to think before responding, compared to in a face-to-face class. These reasons thus seemed to be less important to our participants than to participants in other studies (e.g., Robinson and Hullinger, 2008 ).

Another finding to emerge from Study 1 was that students' preference for one modality over the other was not related to their marks on the related tests, nor to their age. Much of the extant research on online learning has focused on students who might begin at a disadvantage because they do not have the skills to interact confidently with the technology (e.g., Zhang and Perris, 2004 ). Further, as noted by Ituma (2011) , it seems that when students do not struggle with the technical requirements of online learning, age is no barrier to the value that they can gain from this modality of study. However, with the increased use of e-learning in mainstream education, online components are part of the university experience for more and more students from a diverse range of backgrounds (e.g., Chen et al., 2010 ; Holley and Oliver, 2010 ), and appropriate technological skills should not be assumed.

The present results serve as a reminder that it is rather simplistic to consider “online learning” as a unitary concept, to be examined on its own or compared wholesale to face-to-face learning. Both modalities of education have multiple aspects, and research focusing on one particular aspect or combination of aspects might reach very different conclusions from research focusing on another aspect or combination of aspects. Our participants liked completing individual written activities online, but preferred to engage in class discussions in person. As noted earlier, this view may have been strongly influenced by the stilted and drawn-out nature of the online discussion that the current participants experienced, which took place mostly asynchronously, rather than in real time. It is clear that the developers of online content need to consider a range of issues in designing the best way to deliver this content to students. Simply providing materials ( Evans et al., 2004 ) or a discussion forum ( Swan, 2002 ) online does not automatically aid learning ( Davies and Graff, 2005 ), making more contributions to an online discussion does not necessarily lead to better academic performance.

In addition to some of the limitations already mentioned, it should be noted that our sample sizes were modest. The participants were all advanced level undergraduates in psychology, who were enrolled in a face-to-face course but who were also accustomed to doing some of their learning online, in a self-directed way. Thus, further work with a wider range of participants would help to ascertain the generalizability of these findings. It should also be noted that we asked students about only a few aspects of online vs. face-to-face learning, in a unit which provided online materials in a relatively limited range of ways (e.g., asynchronous discussion forum, scrollable written exercises). This specificity is both a strength and a limitation of the current research. Although they provide good experimental control and strong ecological validity, the focused nature of these tasks and their context means that it is important not to overgeneralize our findings. Future research should aim to extend questions about online learning to a wider range of disciplines, using online activities in a more varied range of ways, in order to build up a broader picture of students' preferences and performance in terms of online and in-class learning. Extending research in a number of dimensions will become increasingly important for understanding how blended learning and flipped classrooms can progress along with the technology that underpins these paradigms.

The provision of online units, and of online components to face-to-face units, is continuing to expand worldwide. Those who are responsible for creating online material for blended units would do well to consider carefully the nature and type of activities they allocate to each modality. In directly comparing the same students' performance and perceptions on in-class vs. online learning, this study confirmed that in these groups, at least, online activities led to similar levels of academic achievement as face-to-face activities. It seems that although students appreciate the flexibility of choosing the time and place to do some activities, they also value the greater engagement provided by discussions that take place face-to-face, rather than face-to-screen. Rather than being seen simply as an alternative modality for delivering academic content, the benefits of online technology should be adapted not only to offer greater flexibility, but to inspire students' engagement and success at university and beyond.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Alexander, S. (2001). E-learning developments and experiences. Educ. Train . 43, 240–248. doi: 10.1108/00400910110399247

CrossRef Full Text | Google Scholar

Anderson, T., and Dron, J. (2011). Three generations of distance education pedagogy. Int. Rev. Res. Open Distance Learn . 12, 80–97. Available online at: http://www.irrodl.org/index.php/irrodl/article/view/890/1663/

Australian Qualifications Framework. (2013). Australian Qualifications Frameworks Council, 2nd Edn . Available online at: http://www.aqf.edu.au/

Balluerka, N., Rodríguez, M., Gorostiaga, A., and Vergara, A. (2008). Development of a questionnaire to evaluate pilot schemes adapting undergraduate courses to the requirements of the European higher education area (EHEA). Eur. Psychol . 13, 222–226. doi: 10.1027/1016-9040.13.3.222

Bell, M., Bush, D., Nicholson, P., O'Brien, D., and Tran, T. (2002). Universities Online. A Survey of Online Education and Services in Australia. 02-A Occasional Paper Series . Canberra: Department of Education, Science, and Training.

Braun, V., and Clarke, V. (2006). Using thematic analysis in psychology. Qual. Res. Psychol . 3, 77–101. doi: 10.1191/1478088706qp063oa

Carini, R. M., Kuh, G. D., and Klein, S. P. (2006). Student engagement and student learning: testing the linkages. Res. High. Educ . 47, 1–29. doi: 10.1007/s11162-005-8150-9

Chen, P.-S., Lambert, A. D., and Guidry, K. R. (2010). Engaging online learners: the impact of web-based learning technology on college student engagement. Comput. Educ . 54, 1222–1232. doi: 10.1016/j.compedu.2009.11.008

Citera, M. (1998). Distributed teamwork: the impact of communication media on influence and decision quality. J. Am. Soc. Inf. Sci . 49, 792–800.

Google Scholar

Conole, G., de Laat, M., Dillon, T., and Darby, J. (2008). “Disruptive technologies”, “pedagogical innovation”: what's new? Findings from an in-depth study of students' use and perception of technology. Comput. Educ . 50, 511–524. doi: 10.1016/j.compedu.2007.09.009

Davies, J., and Graff, M. (2005). Performance in e-learning: online participation and student grades. Br. J. Educ. Technol . 36, 657–663. doi: 10.1111/j.1467-8535.2005.00542.x

Evans, C., Gibbons, N. J., Shah, K., and Griffin, D. K. (2004). Virtual learning in the biological sciences: pitfalls of simply “putting notes on the web.” Comput. Educ . 43, 49–61. doi: 10.1016/j.compedu.2003.12.004

Galway, L. P., Corbett, K. K., Takaro, T. K., Tairyan, K., and Frank, E. (2014). A novel integration of online and flipped classroom instructional models in public health higher education. BMC Med. Educ . 14:181. doi: 10.1186/1472-6920-14-181

Pubmed Abstract | Pubmed Full Text | CrossRef Full Text | Google Scholar

Garrison, D. R. (2012). E-learning in the 21st Century: A Framework for Research and Practice . New York, NY: Routledge.

Grieve, R., Indian, M., Witteveen, K., Tolan, G. A., and Marrington, J. (2013). Face-to-face or facebook: can social connectedness be derived online? Comput. Hum. Behav . 29, 605–609. doi: 10.1016/j.chb.2012.11.017

Harden, R. M., and Crosby, J. (2000). AMEE guide no 20: the good teacher is more than a lecturer: the twelve roles of the teacher. Med. Teach . 22, 334–347. doi: 10.1080/014215900409429

Hobbs, D. (2002). Constructivist approach to web course design: a review of the literature. Int. J. E-Learn . 1, 60–65. Available online at: http://www.editlib.org/p/10821

Holley, D., and Oliver, M. (2010). Student engagement and blended learning: portraits of risk. Comput. Educ . 54, 693–700. doi: 10.1016/j.compedu.2009.08.035

Imel, S. (2002). E-learning—Trends and Issues Alert. (Report No-40) . Washington, DC: Office of Educational Research and Improvement.

Ituma, A. (2011). An evaluation of students' perceptions and engagement with e-learning components in a campus based university. Active Learn. High. Educ . 12, 57–68. doi: 10.1177/1469787410387722

Lowenfeld, V. (1939). The Nature of Creative Activity . Oxford: Harcourt Brace.

O'Neill, K., Singh, G., and O'Donoghue, J. (2004). Implementing elearning programmes for higher education: a review of the literature. J. Inf. Technol. Educ . 3, 313–323. Available online at: http://jite.org/documents/Vol3/v3p313-323-131.pdf

Otter, R. R., Seipel, S., Graeff, T., Alexander, B., Boraiko, C., Gray, J., Sadler, K., et al. (2013). Comparing student and faculty perceptions of online and traditional courses. Internet High. Educ . 19, 27–35. doi: 10.1016/j.iheduc.2013.08.001

Piaget, J. (1954). The Construction of Reality in the Child . New York, NY: Basic Books.

Prosser, M., Martin, E., Trigwell, K., Ramsden, P., and Lueckenhausen, G. (2005). Academics' experiences of understanding of their subject matter and the relationships of this to their experiences of teaching and learning. Instr. Sci . 33, 137–157. doi: 10.1007/s11251-004-7687-x

Ramsden, P. (1992). Learning to Teach in Higher Education . London: Routledge.

Robinson, C. C., and Hullinger, H. (2008). New benchmarks in higher education: Student engagement in online learning. J. Educ. Bus . 84, 101–109. doi: 10.3200/JOEB.84.2.101-109

Rossman, M. H. (1999). Successful online teaching using an asynchronous learner discussion forum. J. Asynchronous Learn. Netw . 3, 1–8.

Smith, D., and Hardaker, G. (2000). e-Learning innovation through the implementation of an Internet supported learning environment. Educ. Technol. Soc . 3, 1–16. Available online at: http://www.ifets.info/journals/3_3/e04.html

Strayer, J. F. (2012). How learning in an inverted classroom influences cooperation, innovation, and task orientation. Learn. Environ. Res . 15, 171–193. doi: 10.1007/s10984-012-9108-4

Swan, K. (2002). Building learning communities in online courses: the importance of interaction. Educ. Commun. Inf . 2, 23–49. doi: 10.1080/1463631022000005016

Tucker, B., Halloran, P., and Price, C. (2013). “Student perceptions of the teaching in online learning: an Australian university case study,” in Research and Development in Higher Education: The Place of Learning and Teaching , Vol. 36, eds S. Frielick, N. Buissink-Smith, P. Wyse, J. Billot, J. Hallas, and E. Whitehead (Auckland: Higher Education Research and Development Society of Australasia), 470–484.

Twigg, C. (2003). Improving learning and reducing costs: new models for online learning. Educause Review , Sept/Oct, 28–38. Available online at: http://net.educause.edu/ir/library/pdf/erm0352.pdf

Upton, D. (2006). Online learning in speech and language therapy: student performance and attitudes. Educ. Health 19, 22–31. doi: 10.1080/13576280500534735

Warschauer, M. (1997). Computer-mediated collaborative learning: theory and practice. Mod. Lang. J . 8, 470–481. doi: 10.2307/328890

Zhang, W.-Y., and Perris, K. (2004). Researching the efficacy of online learning: a collaborative effort amongst scholars in Asian open universities. Open Learn . 19, 247–264. doi: 10.1080/0268051042000280110

Keywords: online learning, e-learning, face-to-face learning, university, discussion forums

Citation: Kemp N and Grieve R (2014) Face-to-face or face-to-screen? Undergraduates' opinions and test performance in classroom vs. online learning. Front. Psychol . 5 :1278. doi: 10.3389/fpsyg.2014.01278

Received: 29 August 2014; Accepted: 22 October 2014; Published online: 12 November 2014.

Reviewed by:

Copyright © 2014 Kemp and Grieve. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Nenagh Kemp, Division of Psychology, School of Medicine, University of Tasmania, Locked Bag 30, Hobart 7001, TAS, Australia e-mail: [email protected]

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

• Search Menu
• Sign in through your institution
• Advance articles
• Editor's Choice
• Supplements
• Author Guidelines
• Submission Site
• Open Access
• About Journal of Public Health
• About the Faculty of Public Health of the Royal Colleges of Physicians of the United Kingdom
• Editorial Board
• Self-Archiving Policy
• Dispatch Dates
• Advertising and Corporate Services
• Journals Career Network
• Journals on Oxford Academic
• Books on Oxford Academic

Article Contents

Acknowledgment, conflict of interest, authors’ contribution.

• < Previous

Face-to-face classes during COVID-19: a call for deliberate and well-planned school health protocols in the Philippine context

• Article contents
• Figures & tables
• Supplementary Data

Philip Joseph D Sarmiento, Cora Lyn T Sarmiento, Rina Lyn B Tolentino, Face-to-face classes during COVID-19: a call for deliberate and well-planned school health protocols in the Philippine context, Journal of Public Health , Volume 43, Issue 2, June 2021, Pages e305–e306, https://doi.org/10.1093/pubmed/fdab006

• Permissions Icon Permissions

Schooling is one of the most affected aspects of human life due to coronavirus disease-2019 (COVID-19) pandemic. In a recent correspondence published, the authors reminded every country of their responsibility to come up with strategies to reopen schools safely. This paper reiterates the adherence of school health protocols as significant in the delivery of face-to-face classes following national and international guidelines in mitigating the effects of COVID-19 pandemic as a public health crisis.

Schooling is one of the most affected aspects of human life due to coronavirus disease-2019 (COVID-19) pandemic. Since the rise and threat of the pandemic, many countries around the world have decided to temporarily close schools that have affected millions of students. 1 Consequently, students who are mostly children have been facing a learning crisis due to the pandemic. 2 In a recent correspondence published in this journal, the authors cited that every country has the responsibility to come up with strategies to reopen schools in a safe manner. 3

In the Philippines, the government’s Department of Education has come up with guidelines to implement online and modular distance learning delivery of instruction. 4 This is to safeguard students from being infected by the disease. However, plans to conduct the pilot implementation of limited face-to-face delivery in low-risk areas of COVID-19 transmission for January 2021 have been approved by the president 5 but later recalled 6 due to the threat of the new strain of COVID-19. Predicaments are raised whether the country is ready to open its schools for students to go for face-to-face learning despite having been one of the longest and strictest lockdowns in the world.

School reopening for face-to-face interactions must be carefully planned to ensure the safety of students as well as teachers and school staff in a staged fashion especially in following physical distancing. 7 , 8 Planning and execution of school health protocols during this pandemic must be supported by the truthful data 9 being given by various institutions. Last 11 December 2020, the World Health Organization (WHO) has published a checklist to support school reopening and the preparation for the possible resurgence of COVID-19. 10 WHO cited that ‘The checklist is aligned with, and builds upon, existing COVID-19-related WHO guidelines and is structured around protective measures related to: 1) hand hygiene and respiratory etiquette; 2) physical distancing; 3) use of masks in schools; 4) environmental cleaning and ventilation; and 5) respecting procedures for isolation of all people with symptoms.’ 10 The checklist helps policymakers and school officials to enhance compliance and adherence to public health protocols in the time of the pandemic. 10

In conclusion, school health protocols in conducting face-to-face classes must be planned carefully following national and international guidelines to ensure that students will be safe or at least mitigate the effects of COVID-19. After all, students’ lives matter as education does to them. That is the responsibility of every government to ensure its fulfillment.

No funding was received from this paper.

The authors declare no conflict of interest in this paper.

All authors contributed to all aspects of the manuscript.

UNESCO . UNESCO’s support: Educational response to COVID-19 . 2020 January 1, 2021]; Available from: https://en.unesco.org/covid19/educationresponse/support .

UN . Policy Brief:The Impact of COVID-19 on children , 2020 , 1 – 17 . https://unsdg.un.org/sites/default/files/2020-04/160420_Covid_Children_Policy_Brief.pdf .

Fernandez-Guzman D , Sangster-Carrasco L , Pinedo-Soria A . COVID-19 in Latin America and the Caribbean: what is known about the status of school reopening? J Pub Health 2020 . https://doi.org/10.1093/pubmed/fdaa205 .

Google Scholar

DEPED . DepEd prepares Self-Learning Modules for education’s new normal . 2020 July 11, 2020 January 1, 2021 ]; Available from: https://www.deped.gov.ph/2020/07/02/deped-prepares-self-learning-modules-for-educations-new-normal/ .

DEPED . Official Statement on the Pilot Implementation of Limited Face-to-Face Classes , 2020 . https://www.deped.gov.ph/2020/12/15/official-statement-on-the-pilot-implementation-of-limited-face-to-face-classes/ .

DEPED . On the recall of the President’s approval of the pilot run of face-to-face classes , 2020 . https://www.deped.gov.ph/2020/12/26/on-the-recall-of-the-presidents-approval-of-the-pilot-run-of-face-to-face-classes/#:∼:text=December%2026%2C%202020%20%E2%80%93%20Secretary%20Leonor,originally%20scheduled%20in%20January%202021 .

Viner RM , Bonell C , Drake L et al.  Reopening schools during the COVID-19 pandemic: governments must balance the uncertainty and risks of reopening schools against the clear harms associated with prolonged closure . Arch Dis Child Manila, Philippines: Department of Education, 2020 . https://doi.org/10.1136/archdischild-2020-319963 .

Sheikh A , Sheikh A , Sheikh Z , Dhami S . Reopening schools after the COVID-19 lockdown . J Glob Health 2020 ; 10 ( 1 ): 010376 .

Sarmiento PJD , Yap JFC , Espinosa KAG et al.  The truth must prevail: citizens’ rights to know the truth during the era of COVID-19 . J Pub Health 2020 ;fdaa240. https://doi.org/10.1093/pubmed/fdaa240 .

WHO . Checklist to support schools re-opening and preparation for COVID-19 resurgences or similar public health crises . 2020 December 11, 2020 January 1, 2021]; Available from : https://www.who.int/publications/i/item/9789240017467 .

Email alerts

Citing articles via.

• Recommend to your Library

Affiliations

• Online ISSN 1741-3850
• Print ISSN 1741-3842
• Copyright © 2024 Faculty of Public Health
• About Oxford Academic
• Publish journals with us
• University press partners
• What we publish
• New features  
• Open access
• Institutional account management
• Rights and permissions
• Get help with access
• Accessibility
• Advertising
• Media enquiries
• Oxford University Press
• Oxford Languages
• University of Oxford

Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide

• Copyright © 2024 Oxford University Press
• Cookie settings
• Cookie policy
• Privacy policy
• Legal notice

This Feature Is Available To Subscribers Only

Sign In or Create an Account

This PDF is available to Subscribers Only

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Oxford University Press - PMC COVID-19 Collection

Face-to-face classes during COVID-19: a call for deliberate and well-planned school health protocols in the Philippine context

Philip joseph d sarmiento.

Christian Living Education Department, Holy Angel University, Angeles 2009, Philippines

Cora Lyn T Sarmiento

Sto. Rosario Elementary School, Department of Education-Schools Division of Angeles City, Angeles 2009, Philippines

Rina Lyn B Tolentino

Angeles Elementary School, Department of Education-Schools Division of Angeles City, Angeles 2009, Philippines

Schooling is one of the most affected aspects of human life due to coronavirus disease-2019 (COVID-19) pandemic. In a recent correspondence published, the authors reminded every country of their responsibility to come up with strategies to reopen schools safely. This paper reiterates the adherence of school health protocols as significant in the delivery of face-to-face classes following national and international guidelines in mitigating the effects of COVID-19 pandemic as a public health crisis.

Schooling is one of the most affected aspects of human life due to coronavirus disease-2019 (COVID-19) pandemic. Since the rise and threat of the pandemic, many countries around the world have decided to temporarily close schools that have affected millions of students. 1 Consequently, students who are mostly children have been facing a learning crisis due to the pandemic. 2 In a recent correspondence published in this journal, the authors cited that every country has the responsibility to come up with strategies to reopen schools in a safe manner. 3

In the Philippines, the government’s Department of Education has come up with guidelines to implement online and modular distance learning delivery of instruction. 4 This is to safeguard students from being infected by the disease. However, plans to conduct the pilot implementation of limited face-to-face delivery in low-risk areas of COVID-19 transmission for January 2021 have been approved by the president 5 but later recalled 6 due to the threat of the new strain of COVID-19. Predicaments are raised whether the country is ready to open its schools for students to go for face-to-face learning despite having been one of the longest and strictest lockdowns in the world.

School reopening for face-to-face interactions must be carefully planned to ensure the safety of students as well as teachers and school staff in a staged fashion especially in following physical distancing. 7 , 8 Planning and execution of school health protocols during this pandemic must be supported by the truthful data 9 being given by various institutions. Last 11 December 2020, the World Health Organization (WHO) has published a checklist to support school reopening and the preparation for the possible resurgence of COVID-19. 10 WHO cited that ‘The checklist is aligned with, and builds upon, existing COVID-19-related WHO guidelines and is structured around protective measures related to: 1) hand hygiene and respiratory etiquette; 2) physical distancing; 3) use of masks in schools; 4) environmental cleaning and ventilation; and 5) respecting procedures for isolation of all people with symptoms.’ 10 The checklist helps policymakers and school officials to enhance compliance and adherence to public health protocols in the time of the pandemic. 10

In conclusion, school health protocols in conducting face-to-face classes must be planned carefully following national and international guidelines to ensure that students will be safe or at least mitigate the effects of COVID-19. After all, students’ lives matter as education does to them. That is the responsibility of every government to ensure its fulfillment.

Acknowledgment

No funding was received from this paper.

Contributor Information

Philip Joseph D Sarmiento, Christian Living Education Department, Holy Angel University, Angeles 2009, Philippines.

Cora Lyn T Sarmiento, Sto. Rosario Elementary School, Department of Education-Schools Division of Angeles City, Angeles 2009, Philippines.

Rina Lyn B Tolentino, Angeles Elementary School, Department of Education-Schools Division of Angeles City, Angeles 2009, Philippines.

Conflict of interest

The authors declare no conflict of interest in this paper.

Authors’ contribution

All authors contributed to all aspects of the manuscript.

• DOI: 10.62517/jhve.202416108
• Corpus ID: 271550484

Research on the Construction of Financial Accounting Courses based on SPOC Model

• Published in Journal of Higher Vocational… 1 January 2024
• Education, Business, Computer Science
• Journal of Higher Vocational Education

Related Papers

Showing 1 through 3 of 0 Related Papers

Continual Face Forgery Detection via Historical Distribution Preserving

• Published: 04 September 2024

Cite this article

• Shen Chen 2 ,
• Taiping Yao 2 ,
• Xiaoshuai Sun 1 ,
• Shouhong Ding 2 &
• Rongrong Ji   ORCID: orcid.org/0000-0001-9163-2932 1  

Face forgery techniques have advanced rapidly and pose serious security threats. Existing face forgery detection methods try to learn generalizable features, but they still fall short of practical application. Additionally, finetuning these methods on historical training data is resource-intensive in terms of time and storage. In this paper, we focus on a novel and challenging problem: Continual Face Forgery Detection (CFFD), which aims to efficiently learn from new forgery attacks without forgetting previous ones. Specifically, we propose a Historical Distribution Preserving (HDP) framework that reserves and preserves the distributions of historical faces. To achieve this, we use universal adversarial perturbation (UAP) to simulate historical forgery distribution, and knowledge distillation to maintain the distribution variation of real faces across different models. We also construct a new benchmark for CFFD with three evaluation protocols. Our extensive experiments on the benchmarks show that our method outperforms the state-of-the-art competitors. Our code is available at https://github.com/skJack/HDP .

This is a preview of subscription content, log in via an institution to check access.

Access this article

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or eBook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime

Price includes VAT (Russian Federation)

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Similar content being viewed by others

Thinking in Patch: Towards Generalizable Forgery Detection with Patch Transformation

Minimum Assumption Reconstruction Attacks: Rise of Security and Privacy Threats Against Face Recognition

Analysis of Adversarial Attacks on Face Verification Systems

Explore related subjects.

• Artificial Intelligence

Data Availibility Statement

The data from four public face forgery detection datasets (i.e., FaceForensics++ (Rossler et al., 2019 ), Celeb-DF (Li et al., 2020 ), DFDC (Dolhansky et al., 2020 ), Wild-Deepfake (Zi et al., 2020 ), Multi-Modal-CelebA-HQ (Xia et al., 2021 )) These datasets have been instrumental in advancing research in the field of face forgery detection and have been widely used by the research community. Such datasets support the findings of this study and are available from third party institutions but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of the above-mentioned third-party institutions. This allows the datasets to be used responsibly and ethically while still allowing the scientific community to build upon the findings of this study and advance the state-of-the-art in face forgery detection.

Afchar, D., Nozick, V., Yamagishi, J., & Echizen, I. (2018). Mesonet: A compact facial video forgery detection network. In Wifs, IEEE (pp. 1–7).

Aljundi, R., Babiloni, F., Elhoseiny, M., Rohrbach, M., & Tuytelaars, T. (2018). Memory aware synapses: Learning what (not) to forget. In ECCV (pp. 139–154).

Aljundi, R., Lin, M., Goujaud, B., & Bengio, Y. (2019). Online continual learning with no task boundaries. arXiv preprint arXiv:1903.08671 .

Buzzega, P., Boschini, M., Porrello, A., Abati, D., & Calderara, S. (2020). Dark experience for general continual learning: A strong, simple baseline. NeurIPS, 33 , 15920–15930.

Google Scholar  

Cha, H., Lee, J., & Shin, J. (2021). Co2l: Contrastive continual learning. In ICCV (pp. 9516–9525).

Chaubey, A., Agrawal, N., Barnwal, K., Guliani, K. K., & Mehta, P. (2020). Universal adversarial perturbations: A survey. arXiv preprint arXiv:2005.08087

Chaudhry, A., Gordo, A., Dokania, P. K., Torr, P., & Lopez-Paz, D. (2020). Using hindsight to anchor past knowledge in continual learning. arXiv preprint arXiv:2002.08165

Chen, Z., Sun, K., Zhou, Z., Lin, X., Sun, X., Cao, L., & Ji, R. (2024). Diffusionface: Towards a comprehensive dataset for diffusion-based face forgery analysis. arXiv preprint arXiv:2403.18471

Chen, S., Yao, T., Chen, Y., Ding, S., Li, J., & Ji, R. (2021). Local relation learning for face forgery detection. In AAAI

Dang, H., Liu, F., Stehouwer, J., Liu, X., & Jain, A. K. (2020). On the detection of digital face manipulation. In CVPR (pp. 5781–5790).

Delange, M., Aljundi, R., Masana, M., Parisot, S., Jia, X., Leonardis, A., Slabaugh, G., & Tuytelaars, T. (2021). A continual learning survey: Defying forgetting in classification tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence .

Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009). Imagenet: A large-scale hierarchical image database. In CVPR. IEEE (pp. 248–255)

Dhariwal, P., & Nichol, A. (2021). Diffusion models beat GANs on image synthesis. Advances in Neural Information Processing Systems, 34 , 8780–8794.

Dolhansky, B., Bitton, J., Pflaum, B., Lu, J., Howes, R., Wang, M., & Ferrer, C. C. (2020) The deepfake detection challenge dataset. arXiv preprint arXiv:2006.07397

Ho, J., Jain, A., & Abbeel, P. (2020). Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems, 33 , 6840–6851.

Hou, S., Pan, X., Loy, C. C., Wang, Z., & Lin, D. (2019). Learning a unified classifier incrementally via rebalancing. In CVPR (pp. 831–839).

Jetley, S., Lord, N., & Torr, P. (2018). With friends like these, who needs adversaries? NeurIPS   31 .

Kim, M., Tariq, S., & Woo, S. S. (2021). Fretal: Generalizing deepfake detection using knowledge distillation and representation learning. In CVPR (pp. 1001–1012).

Kirkpatrick, J., Pascanu, R., Rabinowitz, N., Veness, J., Desjardins, G., Rusu, A. A., Milan, K., Quan, J., Ramalho, T., & Grabska-Barwinska, A. (2017). Overcoming catastrophic forgetting in neural networks. Proceedings of the National Academy of Sciences, 114 (13), 3521–3526.

Article   MathSciNet   Google Scholar  

Li, L., Bao, J., Zhang, T., Yang, H., Chen, D., Wen, F., & Guo, B. (2020) Face x-ray for more general face forgery detection. In CVPR , pp. 5001–5010.

Li, C., Huang, Z., Paudel, D.P., Wang, Y., Shahbazi, M., Hong, X., & Van Gool, L. (2023). A continual deepfake detection benchmark: Dataset, methods, and essentials. In Proceedings of the IEEE/CVF winter conference on applications of computer vision (pp. 1339–1349).

Li, J., Wang, Y., Wang, C., Tai, Y., Qian, J., Yang, J., Wang, C., Li, J., & Huang, F. (2019). DSFD: Dual shot face detector. In CVPR , pp. 5060–5069.

Li, Y., Yang, X., Sun, P., Qi, H., & Lyu, S. (2020). Celeb-DF: A large-scale challenging dataset for deepfake forensics. In CVPR pp. 3207–3216.

Li, Z., & Hoiem, D. (2017). Learning without forgetting. IEEE Transactions on Pattern Analysis and Machine Intelligence, 40 (12), 2935–2947.

Article   Google Scholar  

Liu, H., Ji, R., Li, J., Zhang, B., Gao, Y., Wu, Y., & Huang, F. (2019). Universal adversarial perturbation via prior driven uncertainty approximation. In ICCV (pp. 2941–2949).

Luo, A., Li, E., Liu, Y., Kang, X., & Wang, Z. J. (2021). A capsule network based approach for detection of audio spoofing attacks. In ICASSP 2021-2021 IEEE international conference on acoustics, speech and signal processing ( ICASSP ) IEEE (pp. 6359–6363)

Luo, Y., Zhang, Y., Yan, J., & Liu, W. (2021). Generalizing face forgery detection with high-frequency features. In CVPR, (pp. 16317–16326)

Luo, A., Kong, C., Huang, J., Hu, Y., Kang, X., & Kot, A. C. (2023). Beyond the prior forgery knowledge: Mining critical clues for general face forgery detection. IEEE Transactions on Information Forensics and Security, 19 , 1168–1182.

Mai, Z., Li, R., Kim, H., & Sanner, S. (2021). Supervised contrastive replay: Revisiting the nearest class mean classifier in online class-incremental continual learning. In: CVPR (pp. 3589–3599).

Matern, F., Riess, C., & Stamminger, M. (2019). Exploiting visual artifacts to expose deepfakes and face manipulations. In WACVW IEEE , (pp. 83–92).

Moosavi-Dezfooli, S.-M., Fawzi, A., & Frossard, P. (2016) Deepfool: A simple and accurate method to fool deep neural networks. In CVPR (pp. 2574–2582).

Moosavi-Dezfooli, S.-M., Fawzi, A., Fawzi, O., & Frossard, P. (2017). Universal adversarial perturbations. In CVPR pp. 1765–1773.

Moosavi-Dezfooli, S.-M., Fawzi, A., Fawzi, O., Frossard, P., & Soatto, S. (2018). Robustness of classifiers to universal perturbations: A geometric perspective. In ICLR .

Qian, Y., Yin, G., Sheng, L., Chen, Z., & Shao, J. (2020). Thinking in frequency: Face forgery detection by mining frequency-aware clues. In ECCV , (pp. 86–103). Springer: Berlin

Rebuffi, S.-A., Kolesnikov, A., Sperl, G., & Lampert, C. H. (2017). ICARL: Incremental classifier and representation learning. In CVPR (pp. 2001–2010)

Rombach, R., Blattmann, A., Lorenz, D., Esser, P., & Ommer, B. (2022). High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition ( CVPR ) (pp. 10684–10695).

Rossler, A., Cozzolino, D., Verdoliva, L., Riess, C., Thies, J., & Nießner, M. (2019). Faceforensics++: Learning to detect manipulated facial images. In ICCV (pp. 1–11).

Rusu, A. A., Rabinowitz, N. C., Desjardins, G., Soyer, H., Kirkpatrick, J., Kavukcuoglu, K., Pascanu, R., & Hadsell, R. (2016). Progressive neural networks. arXiv preprint arXiv:1606.04671

Shi, W., & Ye, M. (2023). Prototype reminiscence and augmented asymmetric knowledge aggregation for non-exemplar class-incremental learning. In Proceedings of the IEEE/CVF international conference on computer vision (pp. 1772–1781).

Stehouwer, J., Dang, H., Liu, F., Liu, X., & Jain, A. (2019). On the detection of digital face manipulation. arXiv preprint arXiv:1910.01717

Sun, K., Chen, S., Yao, T., Sun, X., Ding, S., & Ji, R. (2023). Towards general visual-linguistic face forgery detection. arXiv preprint arXiv:2307.16545

Sun, K., Liu, H., Yao, T., Sun, X., Chen, S., Ding, S., & Ji, R. (2022). An information theoretic approach for attention-driven face forgery detection. In: ECCV (pp. 111–127). Springer: Berlin.

Sun, K., Liu, H., Ye, Q., Liu, J., Gao, Y., Shao, L., & Ji, R. (2021). Domain general face forgery detection by learning to weight. In AAAI (Vol. 35, pp. 2638–2646).

Sun, K., Yao, T., Chen, S., Ding, S., & Ji, R. (2022) Dual contrastive learning for general face forgery detection. In: AAAI

Szegedy, C., Zaremba, W., Sutskever, I., Bruna, J., Erhan, D., Goodfellow, I., & Fergus, R. (2013). Intriguing properties of neural networks. arXiv preprint arXiv:1312.6199

Tan, M., & Le, Q. V. (2019). Efficientnet: Rethinking model scaling for convolutional neural networks. In ICML .

Thies, J., Zollhöfer, M., Nießner, M., Valgaerts, L., Stamminger, M., & Theobalt, C. (2015). Real-time expression transfer for facial reenactment. ACM Trans. Graph., 34 (6), 183–1.

Tung, F., & Mori, G. (2019). Similarity-preserving knowledge distillation. In ICCV (pp. 1365–1374).

Verma, V. K., Liang, K. J., Mehta, N., Rai, P., & Carin, L. (2021). Efficient feature transformations for discriminative and generative continual learning. In CVPR , pp. 13865–13875

Wang, Z., Zhang, Z., Lee, C.-Y., Zhang, H., Sun, R., Ren, X., Su, G., Perot, V., Dy, J., & Pfister, T. (2021). Learning to prompt for continual learning. arXiv preprint arXiv:2112.08654

Wang, F.-Y., Zhou, D.-W., Ye, H.-J., Zhan, D.-C. (2022). Foster: Feature boosting and compression for class-incremental learning. In: European Conference on Computer Vision (pp. 398–414). Springer: Berlin

Wu, Y., Chen, Y., Wang, L., Ye, Y., Liu, Z., Guo, Y., & Fu, Y. (2019). Large scale incremental learning. In CVPR (pp. 374–382).

Xia, W., Yang, Y., Xue, J.-H., & Wu, B. (2021). Tedigan: Text-guided diverse face image generation and manipulation. In IEEE conference on computer vision and pattern recognition ( CVPR ).

Xu, J., & Zhu, Z. (2018). Reinforced continual learning. NeurIPS   31 .

Yan, S., Xie, J., & He, X. (2021). Der: Dynamically expandable representation for class incremental learning. In CVPR (pp. 3014–3023).

Yan, Z., Zhang, Y., Fan, Y., & Wu, B. (2023). UCF: Uncovering common features for generalizable deepfake detection. In ICCV .

Yang, X., Li, Y., & Lyu, S. (2019). Exposing deep fakes using inconsistent head poses. In ICASSP, IEEE (pp. 8261–8265).

Zając, M., Tuytelaars, T., & Ven, G. M. (2024). Prediction error-based classification for class-incremental learning. In ICLR .

Zhang, C., Benz, P., Imtiaz, T., & Kweon, I. S. (2020). Understanding adversarial examples from the mutual influence of images and perturbations. In CVPR , (pp. 14521–14530).

Zhao, W., Rao, Y., Shi, W., Liu, Z., Zhou, J., & Lu, J. (2023). Diffswap: High-fidelity and controllable face swapping via 3d-aware masked diffusion. In CVPR .

Zhao, T., Xu, X., Xu, M., Ding, H., Xiong, Y., & Xia, W. (2021). Learning self-consistency for deepfake detection. In CVPR , pp. 15023–15033.

Zhao, H., Zhou, W., Chen, D., Wei, T., Zhang, W., & Yu, N. (2021). Multi-attentional deepfake detection. In CVPR .

Zhuang, W., Chu, Q., Tan, Z., Liu, Q., Yuan, H., Miao, C., Luo, Z., & Yu, N. (2022) UIA-VIT: Unsupervised inconsistency-aware method based on vision transformer for face forgery detection. In: European Conference on Computer Vision, pp. 391–407. Springer: Berlin

Zi, B., Chang, M., Chen, J., Ma, X., & Jiang, Y.-G. (2020). Wilddeepfake: A challenging real-world dataset for deepfake detection. In ACM MM (pp. 2382–2390).

Download references

Acknowledgements

This work was partially supported by National Science and Technology Major Project (No. 2022ZD0118202), the National Science Fund for Distinguished Young Scholars (No. 62025603), the National Natural Science Foundation of China (Nos. U21B2037, U22B2051, 62176222, 62176223, 62176226, 62072386, 62072387, 62072389, 62002305 and 62272401), and the Natural Science Foundation of Fujian Province of China (Nos. 2021J01002 and 2022J06001).

Author information

Authors and affiliations.

Key Laboratory of Multimedia Trusted Perception and Efficient Computing, Ministry of Education of China, Xiamen University, Xiamen, Fujian, China

Ke Sun, Xiaoshuai Sun & Rongrong Ji

Youtu Lab, Tencent, China

Shen Chen, Taiping Yao & Shouhong Ding

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Rongrong Ji .

Additional information

Communicated by Segio Escalera.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix A: Analysis of UAP Pseudo Forgery Samples

Additional Visualization of UAP We also provide more visualization results of UAP for the different stages of protocol1 and protocol2. Furthermore, we have added the visualization of UAPs corresponding to AIGC-related methods in the third row. We can observe from Fig.  5 that different stages have their own discriminative UAP patterns.

Analysis the Distribution of Pseudo Forgery Samples Apparently, our strategy relies on a hypothesis that the UAP based pseudo forgery samples should have the same distribution as the actual forgery samples. To verify this conjecture, we illustrate the feature distributions of four well-trained face forgery detection models. Specifically, we first train each model on subsets of FFpp (Rossler et al., 2019 ) i.e.Deepfakes, Face2Face, FaceSwap and NeuralTextures using Eq. 1. Then we generate respective UAP under the guidance of reserve mechanism details in Sect.  3.4 . Finally, we present the feature distribution of real, forgery and pseudo forgery by t-SNE, as shown in Fig.  6 . We can observe that the distribution of pseudo-forged samples almost overlaps with the forgery features across all four datasets, which demonstrates the high similarity between the two distributions. To explain, as stated in previous works (Zhang et al., 2020 ; Jetley et al., 2018 ), UAP are features that contains fixed directional perturbations across class boundaries irrespective of the diversity in their individual appearance. Jetley et al. ( 2018 ) also demonstrate the close relationship between certain directions and class identity, i.e. the network identifies the certain directions as features associated with class identities. Therefore, we argue that the pseudo-forged samples of UAP can maintain the distribution of the forgery class and it is intuitive to use it as the surrogate of actual forgery feature distribution.

UAPs visualization for different attack methods. The first two rows represent traditional deepfake methods, while the last row corresponds to UAPs generated by AIGC methods

Visualization of the feature distribution of four well-trained models via t-SNE. The red dot represents the real features. The green triangle is the corresponding actual forgery features. The blue cross denotes the pseudo forgery features based UAP (Color figure online)

Appendix B: Additional Experimental Results

Result on Protocol 1 with Order2 Since the stage order is agnostic for lifelong learning, we report the details of protocol 1 with order2 , that is NT \(\rightarrow \) DF \(\rightarrow \) F2F \(\rightarrow \) FS . The results are shown in Tab.  11 . Our proposed HDP significantly outperforms all comparing methods in terms of both AVG and PRE. Compared with baseline SFT, our method achieve 25% performance gain in PRE due to the historical distribution preservation. Furthermore, compared with the latest replayed methods such as Co2L and FReTAL required additional buffer size, our method alleviates the catastrophic forgetting by reserving only one UAP per stage. The SOTA results on both order1 and order2 demonstrate our method is insensitive to the stage order.

Result on Protocol 1 with AIGC To expore the generalizability of our method against the latest manipulation methods, we created a continual benchmark related to face AIGC using the latest Diffusion (Ho et al., 2020 ) technology. Specifically, we use DiffusionFace (Chen et al., 2024 ) dataset which employed the Multi-Modal-CelebA-HQ (Xia et al., 2021 ) dataset as the real face data and generated corresponding fake faces using four advanced Diffusion-based generation methods: ADM (Dhariwal & Nichol, 2021 ), DiffSwap (Zhao et al., 2023 ), LDM (Rombach et al., 2022 ), and StableDiffusion 1.5. We then evaluated all the compared methods following Protocol 1. The quantitative results in Tab.  12 show that although the in-domain performance is high, the SFT suffers from severe catastrophic forgetting, with a higher forgetting rate than traditional Deepfake methods, achieving a PRE of 39.17% in terms of ACC. In contrast, our HDP method can improve the PRE by 22% in terms of ACC, demonstrating the scalability and robustness of our approach when faced with such new attacks. Furthermore, compared to other continual learning methods, our method exhibits significant advantages. Specifically, our method achieves a 4% improvement in AVG compared to the current state-of-the-art PEC method. Moreover, our method attains SOTA performance without using a memory bank, unlike other methods that require one, proving the effectiveness of UAP as a feature. Figure  5 illustrates the visualization of the UAP generated from the AIGC dataset. This experiment demonstrates the generalizability of our method when confronted with the most advanced forgery techniques available today.

Result on Protocol 1 with Unseen Metric In addition to the CFFD setting, which can alleviate the poor generalization issue of current methods to a certain extent, our HDP framework also provides performance gains for unseen attacks. To quantify this, we introduce the Unseen metric, which represents the average performance of the model on attack methods that have not been encountered during the training of each stage in Protocol 1. Specifically, for each stage, we calculate the model’s performance on the attack methods that will be introduced in future stages and then average these values across all stages. This metric effectively captures the model’s ability to generalize to unseen attack methods. Tab.  13 presents the Unseen metric for our method and other comparative continual learning methods. Compared to the baseline SFT method, our HDP achieves a 4.49% improvement in the Unseen AUC metric. Furthermore, our method obtains state-of-the-art results compared to other continual learning methods. This can be attributed to our approach of using UAP for data augmentation when training on current attack data, which enhances the network’s ability to learn a broader forgery distribution and capture more essential forgery features. This, in turn, promotes the model’s generalization to unseen attacks. These results further demonstrate the effectiveness of our HDP method against adaptive adversaries and in scenarios with undefined attack vectors.

Appendix C: Protocols Details

In this section, we elucidate the composition of sub-datasets for the three proposed protocols within the Continual Face Forgery Detection (CFFD). Protocol 1 comprises four subsets from FFpp, which are specifically curated to simulate the ever-evolving new attack methods. In contrast, Protocol 2 encompasses four separate face forgery datasets. These datasets introduce substantial domain variations for both authentic and counterfeit images across each phase. This structure results in a richer assortment of attack types at every stage compared to Protocol 1, which focuses on a singular attack type in each phase. Meanwhile, Protocol 3 is formulated to gauge the performance of models when confronted with prolonged sequences of emerging genuine and doctored images. This protocol is segmented into 10 stages, integrating four subsets from FFpp along with evenly segmented versions of the Celeb-DF, Wild-Deepfake, and DFDC datasets. The intricate design of Protocol 3 accentuates the challenge posed by the catastrophic forgetting phenomenon. Through a deeper comprehension of these protocols’ distinct compositions and objectives, researchers are better equipped to devise and assess face forgery detection mechanisms.

Appendix D: Practicals of CFFD

Difference Among Continual Learning, Domain Generalization and Few-Shot Learning In the realm of face forgery detection, several research directions have emerged that bear similarities to continual learning, notably domain generalization and few-shot learning. This section delineates the distinctions between Continual Face Forgery Detection (CFFD) and these methodologies. Few-shot learning aims to optimize a model’s performance on a test set using a minimal amount of training data. The primary focus here is to leverage this scant data to its fullest potential. Domain generalization, on the other hand, trains models using a known source domain and then evaluates their performance on an entirely unseen target domain. The objective is to enhance the model’s capacity to distinguish data it has never encountered before. Continual learning has a distinct goal: to incorporate and adapt to new data streams without erasing or overshadowing previously acquired knowledge. This approach is particularly concerned with avoiding “catastrophic forgetting” when updating with ample new training data.

Exploring the synergies and integrations between CFFD, domain generalization, and few-shot learning promises to be a captivating avenue for future investigations. Domain generalization focuses on learning representations that are robust and transferable across different domains, while few-shot learning aims to quickly adapt to new tasks with limited training examples. Combining these approaches with CFFD could lead to the development of more versatile and adaptable face forgery detection systems. For instance, by leveraging domain generalization techniques, CFFD models could be trained to capture more domain-invariant features, enabling them to generalize better to unseen forgery techniques. Similarly, incorporating few-shot learning strategies could allow CFFD models to rapidly adapt to new forgery methods with only a few examples, reducing the need for extensive retraining.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Sun, K., Chen, S., Yao, T. et al. Continual Face Forgery Detection via Historical Distribution Preserving. Int J Comput Vis (2024). https://doi.org/10.1007/s11263-024-02160-1

Download citation

Received : 12 September 2023

Accepted : 13 June 2024

Published : 04 September 2024

DOI : https://doi.org/10.1007/s11263-024-02160-1

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Face forgery detection
• Continual learning
• Universal adversarial perturbation

Advertisement

• Find a journal
• Publish with us
• Track your research