With Generative Artificial Intelligence (GenAI) tools, the average person can now create artificial (i.e., synthetic) images, videos, and audio clips relatively quickly, cheaply, and with limited source materials (Rana et al., 2022). While these tools offer new creative possibilities, they have also contributed to the rise of deepfakes: AI-generated synthetic media in which a real person’s likeness or voice is appropriated and altered, often without their consent (e.g., Ali et al., 2021). For example, viral AI-generated images of actor Zendaya getting married recently circulated online as if they were authentic photographs, illustrating how synthetic images can be realistic enough to be misleading (Glynn, 2026; see the images at www.bbc.com/news/articles/clyz7llp4k2o).
While not all deepfakes are created for harm (e.g., fan edits of cartoon characters; Foley, 2024; Pandey et al., 2021), our study focused on cases in which people create or share deepfakes to mislead, misrepresent, harass, bully, or shame others. As people create and share deepfakes of others without their consent, this raises important questions about privacy, consent, and how someone’s image can travel far beyond the context in which they originally shared it (Nissenbaum, 2010).
Prior research on cyberbullying and online harassment has identified that a common and harmful practice among today’s hyperconnected youth is the appropriation of an individual’s image without their consent (Alexander, 2025; Banerji, 2025; Byrne, 2020; Byrne et al., 2023; Byrne & Hollingsworth, 2025; Thorn, 2025). Although deepfake technology was once expensive, time consuming, and only available to experts, it has recently become fast, affordable, and accessible to the average user, including youth (Cover, 2022). Our study explored this emerging phenomenon in the daily lives of high school teachers and asked, “What are teachers’ recommendations for policy, practice, and design for mitigating these deepfake-related harms?”
To answer this research question, we conducted a codesign study (Guha et al., 2013), in which we partnered with nine public high school teachers to brainstorm, design, and respond to realistic scenarios of deepfakes in high school contexts. We selected a codesign approach because deepfake-related harms are context-dependent and rapidly evolving. This approach allowed teachers to draw on their lived experiences to generate realistic scenarios and identify gaps in policy, practice, and design of educational and technological supports.
Our study design was intentionally created to address three aims. First, we observed how teachers with limited prior familiarity with deepfakes could develop a working understanding through collaborative, scenario-based discussion. Second, we codesigned realistic deepfake scenarios grounded in school contexts to surface gaps in policy, practice, and the design of technological and educational supports. Third, we used these newly designed realistic scenarios as case studies to solicit teachers’ responses and recommendations for addressing deepfake-related harms. Like Wei et al. (2025), we explored high school teachers’ perspectives on deepfakes to open a dialog about the possible impact of this new technology on teachers and the implications for both their practice and professional development.
Related Literature
Although many scholars are now researching GenAI use in teaching and learning, there has been little attention to the use of GenAI to create harmful deepfakes in K-12 contexts (Alexander, 2025; Wei et al., 2025). Our study is among the first to examine this emerging phenomenon. While the technologies involved may be novel, the underlying behavior (i.e., nonconsensual manipulation and sharing of a person’s likeness) is not new (Byrne 2020). Deepfakes represent a technology-enhanced form of harassment, in which individuals capture, alter, and share another person’s image without consent. Contemporary GenAI tools now allow adolescents to produce harmful, violent, or sexualized deepfake images and videos of peers with minimal source material and limited technical skill. This capacity raises significant concerns regarding digital ethics, consent, and technology-facilitated harassment. A salient example occurred at a private school in Lancaster, Pennsylvania, where two 14 year old boys generated and disseminated over 350 deepfake sexual images of more than 50 female classmates (Wilson & Doran, 2024). After a lengthy legal battle, the boys were sentenced to probation and 60 hours of community service (Scolforo, 2026).
Like other forms of online harassment, deepfake-related victimization can be traumatic and can lead to significant social, emotional, and academic consequences (Holt et al. 2014). As a result, victims have reported experiencing mental health challenges, difficulty engaging in school or social events, withdrawing from online spaces, and experiencing a form of hypervigilance about their photo being taken (Alipan et al., 2018; Marder et al., 2016). The prevalence of online harassment has led to anticyberbullying education in schools, including teacher professional development, bystander intervention programs, antibullying policies and laws, and reporting procedures within social media platforms (Bauman & Baldasare; 2015; Byers & Cerulli, 2021; Smith & Yoon, 2013). It remains unclear how these educational programs and policies have prepared teachers and school administrators to respond to deepfakes used for interpersonal harm (Wei et al., 2025).
The existing research on deepfakes in high school contexts has primarily focused on AI literacy (Ng et al., 2021a, b) and navigating misinformation online (Hwang et al., 2021; Wineburg & McGrew, 2019). From our review, popular media and journalism are leading the investigation into the use of deepfakes for interpersonal harm in high school contexts (e.g., Ray, 2023). For example, journalists have written about how teens have used GenAI to create deepfake sexual images of other youth (Burgess, 2026; Elliott, 2024; Singer, 2024a, b), as well as a case of how a disgruntled teacher used GenAI to create a deepfake targeting his principal (Diaz, 2024). Singer (2024a), who covers teen privacy for the New York Times, referred to this modern phenomenon as “an epidemic of deepfake nudes in schools” (p. 1). Such cases demonstrate the potential harms that deepfakes can cause to a school community. Our study addressed the gap in the scholarly literature and explored the emerging phenomenon of deepfakes in school contexts.
Deepfakes
A type of synthetic AI-generated media, a “deepfake” is a portmanteau combining deep learning and fake (Rana et al., 2022). The term was coined by a Reddit user in 2017 to describe how they had used deep learning methods to swap the faces of women in pornographic videos with the faces of other people, including famous actresses, to create realistic pornographic videos without people’s consent (Knight, 2018; Trammel, 2023). Since 2017, advances in deep learning and generative adversarial networks have made it possible to produce these nonconsensual, believable deepfakes faster and in a manner that is less detectable to the human eye (Rana et al., 2022).
Today there are many low-cost desktop and mobile applications (apps) for users to create deepfakes and synthetic sexual images quickly with limited source materials (Singer, 2024a, b). While some of these apps are simple face-swapping apps targeted at youth (e.g., Reface, 2025), many others are “nudification” apps that use AI to create synthetic, nonconsensual sexually explicit images of real people (Burgess, 2026; Gibson et al., 2024).
AI Literacy
Our study was informed by the ways people develop digital, media, and technology literacies (e.g., Kumar & Byrne, 2022). While literacy is often defined in everyday discourse as mastery of a cognitive process, we take a broader view. Literacy is a complex social practice in which a person develops knowledge, skills, and abilities through real-world experiences, and often with others (Gee, 2015). We approach AI literacy in the same way.
AI literacy is not simply knowing facts about AI or how to use a tool. Rather, it is the ability to understand, evaluate, detect, and use AI in critical, appropriate, and ethical ways across real-world contexts. Prior research similarly defines AI literacy as “a set of competencies that [enable] individuals to critically evaluate AI technologies, communicate and collaborate effectively with AI, and use AI as a tool online, at home, and in the workplace” (Long et al., 2022, p. 1). In this view, an AI-literate person has the knowledge, skills, and self-efficacy to question and appropriately employ AI in relevant personal or professional situations (Chiu et al., 2024; Laupichler et al., 2022, Mills et al., 2024, Wang et al., 2023).
Since the public release of ChatGPT in 2022, research on fostering AI literacy among teachers has grown quickly. Yet this body of work is relatively new and uneven, with most studies focused on preservice teachers and GenAI use in instructional settings (e.g., Le et al., 2026). Increasingly, this literature asks not only whether teachers can use AI, but whether they can use it in ways that are pedagogically meaningful, ethically responsible, and contextually appropriate (e.g., Kennedy & Castek, 2025; Tagare et al., 2025; U.S. Department of Education, Office of Educational Technology, 2024). Emerging findings suggest that AI literacy is best fostered through active, applied, and collaborative learning experiences rather than one-way technical training (e.g., Dilek et al., 2025). Still, little of this work has examined how in-service teachers develop AI literacy beyond curriculum design and classroom instruction.
Even novice AI users have some existing knowledge and experience with the privacy and security norms of online media. Supporting teachers in becoming more AI literate, therefore, requires respecting and building upon this existing knowledge and helping them connect new concepts to their lived experiences (e.g., Freire et al., 2018). We adopt this approach in our work with high school teachers who initially identified as unfamiliar with deepfakes and synthetic media.
One of the ways in which people develop AI literacy and self-efficacy is through active learning with the appropriate scaffolding (e.g., Hmelo et al., 2000; Hmelo-Silver et al., 2007). Considering AI literacy development using Vygotsky’s (1978) Zone of Proximal Development, people can accomplish more complex tasks with appropriate scaffolding than they could independently. Scaffolding intentionally offloads some of the more complex or tedious aspects of a task so that learners can concentrate and make progress (Reiser & Tabak, 2014). Examples of scaffolding include just-in-time expert feedback, guided facilitation, and worked examples. With these scaffolds, learners are more able to engage with the content of a problem, such as a realistic scenario or case study that matches their level of expertise.
Methods
We present findings from an IRB-approved, exploratory qualitative pilot study conducted at a public Historically Black College or University (HBCU) in the MidAtlantic US and funded, in part, by the National Science Foundation Institute for Trustworthy AI in Law & Society (TRAILS; Award No. 2229885). We used a codesign session as our primary data collection activity. Codesign refers to a facilitated participatory process in which researchers and stakeholders collaborate to address a problem by brainstorming ideas, scenarios, and design recommendations grounded in the participants’ lived experiences (Guha et al., 2013; Penuel et al., 2007). We use the term codesign to describe that session and our general participatory approach to data collection, not to suggest that our project was an intervention or a design-based research study.
Data Collection
In summer 2024, 14 high school teachers were participating in an unrelated AI enrichment program hosted at the same HBCU. We invited these teachers to participate in our 2-hour, in-person codesign session about deepfakes in high school contexts (hereinafter “our session”). Prior to our session, teachers completed an online form that collected their demographic information and informed consent. Out of the 14 teachers we recruited, nine participated in our 2-hour session.
The session was facilitated by the first two authors (see Table 1). Sitting in a large circle, we asked all participants to introduce themselves including their interest in being at our session. Then, we facilitated a 15-minute whole-group discussion about deepfakes to establish a shared understanding about the term. We asked questions like, “What comes to mind when you think of deepfakes? What are some examples?”
Table 1
Session Overview
| Task | Main Points & Prompts | Duration |
|---|---|---|
| Icebreaker | Facilitators provide an overview of the consent and codesign process. Then, each participant introduces themselves. | 5 minutes |
| Large Group Discussion | Facilitator Prompts: - What comes to mind when you think of deepfakes? - Where do you hear about deepfakes? - What are some examples of deepfakes that you have heard about or experienced? | 15 minutes |
| Small Groups Generate Scenarios | Draft a realistic scenario of a deepfake from your context including the response from all stakeholders…Write this scenario on a flipchart paper. Then, on a second flipchart paper, pose recommendations for how this scenario could have been mitigated or handled better. For example, What are your recommendations to administrators, lawmakers, and AI designers? What tools or resources do you wish you had? What education do you wish you and your colleagues had had? | 45 minutes |
| Small Groups Swap Scenarios | Swap scenarios with the other group. Read and discuss the other group’s scenario. Then, on a new flipchart paper, pose recommendations for how this scenario could have been mitigated. What are your recommendations to administrators, lawmakers, and AI designers? What tools or resources do you wish you had? What education do you wish you and your colleagues had had? | 15 minutes |
| Large Group Discussion | Discuss each scenario as a large group, allowing each group time to share their written recommendations. - Facilitators identify commonalities between each small group’s recommendations and ask clarifying questions. - Facilitators articulate a synthesis of recommendations. Final Prompt: What are your final thoughts or takeaways about the possible impact of deepfakes in their high school contexts? | 45 minutes |
We then split into two groups with four teachers and one researcher in one group and five teachers and one researcher in the other. We gave each group flipchart paper and markers. Over the next 45 minutes, we asked each group to draft a realistic scenario of a deepfake in their school context. Then, on a second flipchart paper, we asked them to write recommendations for administrators, lawmakers, and AI developers in response to that scenario. The researchers facilitated the conversations to answer questions about deepfake technologies and relevant policies and to refocus the discussion on the prompt, when needed. We also took notes about the conversation, which we saved as field notes.
Next, we asked the groups to swap their scenarios. Each group had 15 minutes to discuss the other group’s scenario and provide recommendations on a new sheet of flipchart paper. In our final 45 minutes, we facilitated a whole-group discussion of both scenarios and asked the participating teachers to share their overall takeaways about the impact of deepfakes in high school contexts. We audio-recorded the session using Zoom and transcribed it using Zoom’s embedded AI.
After the session, the first two authors, who had served as facilitators, met to debrief. Together, we transcribed the flipchart artifacts and our field notes into a report consisting of the two scenarios and a synthesis of the groups’ responses to each scenario. As a form of member checking, we emailed this postsession report to the participants and asked for feedback or edits within 1 week. Participants either agreed that our summarization was accurate (N = 1) or provided no response (N = 8).
Participants
All nine participating teachers work for a local public school district in the MidAtlantic US, either urban or suburban. As presented in Table 2, the majority of participants identified as Black or African American (N = 5) and women (N = 6). The group spanned from a teacher who had just completed her 1st-year teaching to a senior teacher with over 22 years of experience; however, most teachers were between their 3rd- and 7th-year teaching. Teachers identified their primary discipline as science (N = 3) and English (N = 3). Because of their interest and prior participation in an AI-related summer research program, we anticipated that these teachers would have more AI literacy and engagement than the average high school teacher. For this reason, we do not argue that these teachers are representative of all high school teachers.
Table 2
Participant Demographics
| Identifier | Race/Ethnicity | Age | Gender | Years Teaching | Primary Discipline |
|---|---|---|---|---|---|
| Teacher A [TA] | Black or African American | 40 - 49 | Woman | 2 years | Science |
| Teacher B [TB] | Black or African American | 20 - 29 | Woman | 4 years | English & ELA |
| Teacher C [TC] | Black or African American | 30 - 39 | Woman | 9 years | Health |
| Teacher D [TD] | White or European American | 30 - 39 | Woman | 8 years | Science |
| Teacher E [TE] | Black or African American | 30 - 39 | Man | 10 years | School Counselor |
| Teacher F [TF] | Black or African American | 20 - 29 | Woman | 1 year | English & ELA |
| Teacher G [TG] | Prefer not to disclose | 40 - 49 | Man | 22 years | Career & Technical Education |
| Teacher H [TH] | White or European American | 30 - 39 | Man | 12 years | Science |
| Teacher I [TI] | Black or African American | 50 - 59 | Man | 21 years | English & ELA |
Data Analysis
The research team includes one tenured faculty member and one Science Education doctoral candidate, who are the first two authors and who facilitated the codesign session, as well as two additional student-researchers (a Science Education doctoral student and a Psychology undergraduate student), who contributed to our analysis. We each brought experience in AI literacy and K-12 education, which informed our interpretation. Because the first two authors facilitated the session and developed the postsession report, we approached the analysis reflexively and treated the transcript, flipcharts, field notes, and report as complementary records of a single codesign session.
We conducted our thematic analysis (Braun & Clarke, 2006; Saldaña, 2013) in two stages to examine how teachers made sense of deepfakes in high school contexts and what recommendations they generated through the codesign session. First, two researchers reviewed the postsession report, the session recording and transcript, flipchart artifacts, and the researcher field notes to generate an initial list of open codes. These codes reflected teachers’ perceptions of and feelings about deepfakes (e.g., worry, uncertainty, fear, anger, hopelessness, or helplessness) and recommendations (e.g., need for training, need for policy, or need for stronger tools). We coded short excerpts, discussion turns, and artifact text that expressed a single idea, reaction, or recommendation.
Then, in our second stage, all four research team members independently reviewed the data using this initial list of codes. We then met to compare interpretations, discuss discrepancies, and refine our codes. Next, we grouped related codes into broader categories and identified patterns across the data sources. We synthesized these categories into themes by identifying the central ideas and what best captured teachers’ perceptions, feelings, and recommendations related to deepfakes in high school settings.
Finally, we discussed the best approach to presenting our findings. In the findings section we refer to teachers by an indicator acronym (e.g., Teacher A is TA) presented in Table 1. Because this was a pilot codesign study, our analysis aimed to identify salient patterns in teachers’ sensemaking and recommendations rather than to generate a formal theory.
Findings
We observed that participants’ knowledge of and attitudes about deepfakes evolved throughout the design session. Thus, we present our findings as they manifested chronologically during the session in alignment with the three aims of the paper. We begin by synthesizing our initial large group discussion about deepfakes in their school contexts and online lives.
Teachers Quickly Gained Working Knowledge and Contextual Awareness
In our opening conversation, several teachers shared that they had not heard the term deepfake prior to our email; however they quickly seemed to understand the terminology and concept when provided with examples from everyday life. The few participants who were aware of the term prior to the session defined a deepfake as “when you have, like, a video that, like, it looks like it’s someone else, but it’s typically like maybe a face put on a different body or something” [TE], “…cloning sound or face…” [TF], and “the technology takes the, like, the audio or the video that you already have…. It learns your patterns to then replicate it” [TF]. Teacher I [TI] then discussed the difference between altered media and AI generated media:
I guess the difference between, I mean, I don’t know. I guess the word “deep fakes,” meaning like, I guess a regular fake is one thing…but a deep fake might be… when you can’t necessarily tell. Where professionals have a little difficulty trying to figure out if it’s real or not, by checking the audio to see if it was recorded over or changed in some kind of way or checking the video to see if it was altered. Assuming “deep,” kinda’ it means that.
Through only a short conversation, all teachers were able to accurately articulate that deepfake and synthetic media are artificially generated based on training data.
With this working definition, almost all the teachers shared that they had encountered synthetic media and deepfakes in popular media. After hearing others’ stories, Teacher D [TD] said, “I didn’t know what deepfakes were before. Like, I didn’t know what that word meant. I didn’t know how common it could be, like, in social media, etcetera.” Additionally, Teacher C [TC] realized that they even had firsthand experience with deepfakes in their school districts: “I just didn’t know the proper terminology. But it all makes sense. I mean, it just recently happened in [a school I know]. So, it all makes sense.”
Regardless of their prior knowledge, all teachers quickly understood that deepfakes can be used to “create biases” [TH], “manipulate situations” [TG], shame someone sexually, or frame someone for saying something harmful. Despite an incoming lack of awareness of the term, the teachers gained a working understanding and recalled examples from their everyday lives within a 15-minute group discussion.
Our discussion then organically moved through stages, beginning with realizing what was possible, to concerns about privacy, to feeling like eventually they will be a deepfake victim, to finding hope and potential positive uses of synthetic media. First, most teachers were unaware that the average technologically skilled person could create a deepfake or synthetic image with limited source material. For example, in talking about a real case of a teacher creating a deepfake of their principal, participants were shocked that the teacher had the technical skills and tools to create a believable audio clip. Wondering aloud, Teacher B said, “How did that teacher make a deepfake audio of that principal? How did he even.… Is he smart? How was he smart enough to do that?” The assumption being that the general user is not likely “smart enough” to create a believable deepfake.
Second, teachers shared concerns about how their media might be used without their consent. Teacher E [TE], who supports students in creating podcasts, shared how he no longer encourages his students to use a certain online video editing tool because of privacy concerns and worries that it will accidentally subject students and teachers to being the victims of deepfakes. He stated,
I realized that when you use [the website] and you upload your videos.… They have the right to use [the videos] on their own. So, I didn’t want to use it.… Like, Wait a minute! This makes me just feel … this just feels wrong. Like if I upload my video … it’s now part of their database. And there’s so many people that use [that website], and probably don’t realize anything you upload to it, it becomes part of its’ net. And I think that’s where a lot of these … deepfake videos get sourced from.
In response to this sharing, other teachers realized that they may have uploaded media of themselves or their students in a way that opens students and teachers to privacy violations. Teachers wrestled with these privacy issues in conversation such as when one teacher asked, “How could people govern their content, or like guard it?” [TF] to which another teacher (TB) responded,
You can’t. If it’s out there, it’s out. There’s nothing you can really do to…. If you make a TikTok, and then it’s spread and viral and you delete it, it’s already on so many different platforms. Even though you delete it off your original platform, people will still have access to it.
Other teachers echoed this sense of hopelessness and resignation that once media has been shared, it cannot be unshared. The teachers worried aloud about the potential misuses of their available media and how it could be altered to harm them and their reputation (e.g., AI generated police bodycam footage). This unprompted conversation revealed that the teachers worry about the misuse of their image and the ramifications now made possible with AI.
In response, teachers sought solutions. Turning to the facilitators, they wanted to know more about AI detection tools and “cyber forensics” [TH] — hoping that there are processes for identifying a deepfake that are as easy and accessible as the tools for creating a deepfake. They wanted more AI literacy training for teachers, school administrators, parents, and community members. As Teacher I [TI] said, “A lot of teachers have no clue [about deepfakes]. I mean, I work with teachers that… they have no clue.… They hear about it, but they don’t know how it works.”
Teachers said that the community needs to have the AI literacy to recognize synthetic media and the psychological readiness to address future deepfake incidents. In other words, the teachers wanted adult-focused AI literacy education so that young people have informed adults around them to be critical consumers of AI and know when AI generated content cannot be trusted.
The teachers concluded this initial whole-group conversation by brainstorming ways to teach students about deepfake technology. For example, Teacher E suggested integrating the topic into social studies, saying, “Maybe my history/social studies teachers, and when they cover stuff like propaganda.” Teacher C similarly connected deepfakes to health class lessons on social media: “I can introduce this terminology to my students, and we can go over scenarios of … deepfakes…And like, what are the laws behind that?” Other teachers extended the conversation by considering how synthetic media could be used productively by students, such as for vision boarding.
When a researcher asked why they were trying to find a positive use case for synthetic media Teacher H said,
Just in the essence of like, we shouldn’t ignore [AI generated media]…. We’ve all had to go through, like, digital awareness as the Internet’s developed and find different platforms. Like, it’s always good as an educator [to be a] step ahead of knowing it and applying [it to] more positive outlets.
Similarly, Teacher B said,
[It’s] like a good thing [students] can deal with [synthetic media]. A lot of [students] won’t even go down the bad things. But if we introduce [synthetic media using an approach like] “why you shouldn’t use this,” because they can do this, this and that, then they’re going to try to [make harmful deepfakes]. But if you just introduce all the good things you can do with [AI], only showing that, very few [students] will go down the other way.
We ended this discussion to move into small groups, but teachers were still writing notes of the positive ways they could introduce their students to deepfakes.
Designing Realistic Deepfake Scenarios and Recommendations
When the small groups convened, both followed a similar pattern. The teachers first discussed their surprise about the reality of deepfakes. Then they expressed consternation and concern that today’s youth have access to much more technology and online connectivity than they had growing up. Then, each group (organically) began discussing how students use AI-generated text for plagiarism and cheating. Each group’s assigned researcher/facilitator then refocused the conversation toward scenarios about AI-generated deepfakes. Once focused on the prompt, each small group was able to create a plausible scenario of deepfakes being used for harm in a high school context.
In this section, we present the full scenarios created by the teacher-researcher small groups, followed by the salient issues, proposed solutions, and recommendations for policymakers, administrators, and researchers. We synthesized the solutions that emerged when each group responded to its own scenario with those that emerged after the groups swapped and responded to the other group’s scenario.
Scenario 1: Fight Video
Student1 is failing Teacher1’s class due to poor attendance. Student1 approached the teacher to change their grade, but Teacher1 refused to do so. Student1 and Student2 decide to record themselves fighting on video. They then altered the video using AI to replace Student2’s face with that of Teacher1. Once completed, the deepfake video depicts Teacher1 and Student1 engaged in a physical altercation in an empty classroom. Student1 then shared the video across their various social media platforms. The video went viral. Teacher1 was placed on immediate leave by district leaders. Parents are in an uproar over the video. Other students in the school begin to spread rumors and instigate arguments with other teachers. Fellow teachers have isolated Teacher1.
In response to Scenario 1, teachers were concerned that members of the school community would not have the AI literacy to know that a student could create such a video and that the teacher involved would be presumed guilty. For this reason, AI literacy education for the community inside and outside of school was unanimously identified as a primary need. Additionally, teachers identified that students need to be taught about the ethics of AI and the expectations for consent when using another person’s image.
Beyond the need for more education and awareness, teachers, particularly veteran teachers, said that administrators need AI detection tools for quickly identifying if media is AI-generated and to identify the creator. The teachers were concerned that the current state of AI detection would take too long for Teacher1 in the scenario to salvage their reputation. They viewed Teacher1 as helpless. These insufficient verification methods were concerning to teachers, because they were aware of the potential for biases to emerge. Further, the teachers believed that these detection tools exist but are not yet accessible to school administrators.
Once a creator is identified, teachers said administrators need policies and procedures to hold the student (and potentially their parents) accountable. They were concerned about the lack of legislation around GenAI and the people’s right to their own image. The teachers empathized with Teacher1 and discussed how the damage to their reputation could be repaired. A teacher asked why a person could not request their image be deleted from the Internet, and we discussed the European Union’s Right to Erasure or Right to be Forgotten policy, an idea many of the teachers wished was applicable in their context. Finally, the teachers discussed restorative practices between Student1 and Teacher1 to remedy the harm such as a public apology or mediation.
Scenario 2: Deepnudes
A deepfake nude (i.e., a deepnude) of a teen student is circulating around various group messenger platforms (e.g., snapchat, iMessage, text threads, Instagram messages, and discord). Teachers in the building did not find out about the image until much later. Some teachers find out from a student informant, and some find out from the victim. By the time teachers find out about the deepnude, it has been widely shared. The victim of this deepnude has had a drastic change in their behavior. In response, the teachers share their observations and information with the school’s administration and student welfare team. A group meeting is held with school admin, school police, and school counselors to better understand the severity of the situation. The victims’ parents are contacted. There is still some uncertainty if the images are real or fake. The district publicist and other media teams begin to get involved and there is a public fallout. There is still no clear indication if this is a real or fake image. Also, the image is now unable to be deleted.
In response to this scenario, teachers acknowledged that, without strong relationships with students and a basic understanding of deepfakes, they might misinterpret this scenario and offer inadequate support to the students. Many expressed surprise and concern that deepfakes can now be created by everyday users, including teenagers. Their concern grew as they discussed both the ease and accessibility of the AI technology and the lack of robust regulation preventing the creation of sexualized deepfakes of children. Across both groups, teachers emphasized the need for AI literacy training for teachers and students, particularly about the legality and ethics about deepfakes. We also discussed teaching students about the difference between online jokes and harm.
Teachers reacted with worry and frustration, echoing the concerns raised in the first scenario about how schools can protect students (and themselves) from the consequences of deepfakes. Several teachers described feeling anxious about their inability to distinguish authentic images from AI-generated ones or to identify the creator of a deepfake. For example, Teacher H expressed that administrators need a reliable way to determine the authenticity of images and identify a deepfake’s creator. Teachers hoped that policies and laws already existed to address this nonconsensual creation of nude images, but they were unaware of any concrete protections. They assumed that their School Resource Officer or local police would understand how to navigate such situations but reported feeling uneasy about their own lack of knowledge. They wished they knew more about the legal protections for people to protect their images online, including the right to delete AI-altered images of themselves.
Teachers also worried that their school’s student code of conduct might not adequately address the misuse of students’ images without consent, including incidents that take place outside of school. They recommended creating clear consequences and a restorative justice process to ensure that creators of deepfakes take accountability for their actions and that the victims get some mediation and closure. Across the discussions, teachers conveyed a strong sense of urgency and emotional discomfort, underscoring their desire for clearer policies, training, and legal guidance to support students experiencing technology-facilitated harm.
Discussion
Our codesign study offers an early exploration of how high school teachers make sense of deepfakes as an emerging form of interpersonal harm in school contexts. In June 2024, two researchers and nine public high school teachers convened for a 2-hour, in-person codesign session to discuss the potential harmful use of deepfakes in a high school setting and design plausible scenarios.
We found that, through an initial group conversation (rather than a lecture or intervention), participating teachers who began the session unaware of deepfakes were able to quickly develop a working understanding the concept and engage in meaningful discussions about relevant AI ethics and policy. We observed that most teachers entered the sessions with minimal AI literacy regarding deepfakes but were able to quickly understand the concept and see its relevance to their everyday online lives. Through conversation, teachers articulated how deepfakes could be used to manipulate or misrepresent people in a way that is unethical or harmful. They also considered how to mitigate harm at multiple levels: as individuals (e.g., avoiding posting their videos on untrustworthy platforms that could use their content to train AI), as educators (e.g., incorporating AI literacy in their curriculum), and as community members (e.g., advocating for AI detection tools and enforceable regulations).
Much of today’s AI-focused professional development for teachers remains heavily content driven, neglecting the processes through which educators build technological literacy and develop self-efficacy in adopting new tools (Byrne et al., 2025). Rather than approaching technology literacy as an isolated skill to be mastered, our findings suggest collaborative dialogues with peers may be a promising pathway for teachers to make sense of new technologies. By exchanging experiences and sharing narratives from their everyday online lives, the teachers in our session developed a deeper, more authentic understanding of deepfakes and their implications.
These kinds of engaging, vicarious experiences have been shown to promote self-efficacy and confidence with new technologies (Tschannen-Moran & McMaster, 2009). This also extends emerging work on AI literacy by suggesting that scenario-based, collaborative dialogue may support teachers’ sensemaking around unfamiliar technologies. From our 2-hour session, we observed that teachers were actively practicing AI literacy and self-efficacy through hands-on and creative engagement.
At the same time, we did not fully anticipate the intensity of teachers’ emotional reactions to the content of our session. Teachers expressed concern, discomfort, and uncertainty about their ability to protect their own images while continuing to participate in social media communities. Our observations suggest that some teachers may have been grappling with a diminished sense of online safety and a loss of control over their own image and voice.
We draw on Nissenbaum’s (2010) theory of contextual integrity as an interpretive lens to better understand teachers’ reactions, particularly in relation to privacy. While Nissenbaum’s theory did not guide our data collection or analysis, it helps explain how teachers’ expectations for the appropriate flow of personal information were disrupted. Many teachers were previously unaware or had not considered how their publicly shared likeness could be used to generate synthetic media. Instead, they tended to trust that everyday online platforms operate transparently and that their images circulate only within expected social boundaries (e.g., Instagram posts were only viewed by intended followers).
Through our session, teachers began to examine critically how those same images and videos could be copied, repurposed, and manipulated into harmful deepfakes without their consent or awareness. From this perspective, deepfakes represent not only a breach of contextual norms, but also a significant privacy violation, as they involve the unauthorized use of personal data outside of the original context. Realizing that malicious actors could access their likeness and use deepfake tools appeared to heighten teachers’ sense of vulnerability and create mistrust in the online platforms they regularly use.
We also observed that teachers were surprised and upset that their typical online sharing practices were exposing them and their students to these risks. In response, some pivoted quickly to action — calling for AI detection tools and AI literacy training or finding positive or pedagogical uses for synthetic media. Our session design did not fully anticipate the emotional impact of learning about the potential privacy risks, particularly the unsettling possibility that one’s image can be appropriated and manipulated without consent. These findings point to the need for more intentional approaches to AI literacy professional development. Programs may need to address not only technical knowledge and ethical guidelines but also the emotional aspect of learning about technology-facilitated privacy harms and to cultivate a sense of agency rather than helplessness among participants.
The codesign process itself played a critical role in surfacing these insights. With limited scaffolding, teachers were able to create realistic scenarios that closely resembled real-world incidents and reflected the complexity of school contexts, including the expected responses from students, administrators, and community members. These scenarios allowed teachers to move beyond abstract concerns and engage with concrete, situated problems. In doing so, theyidentified gaps in existing policy, limitations of available tools, and challenges in coordinating stakeholders. This aligns with prior codesign research demonstrating that participants bring critical contextual knowledge that can surface problems and solutions that might otherwise be overlooked (Guha et al., 2013; Penuel, 2016). We presented these scenarios for teachers, administrators, researchers, policymakers, and teacher educators to consider as deepfake case studies that are salient to teachers and require action.
Finally, we found that the use of these newly designed, realistic scenarios enabled teachers to engage quickly and substantively in discussions on policy, practice, and design. In particular, teachers had meaningful conversations about AI ethics, AI policy at all levels (i.e., school, district, state, and federal), classroom- and school-level procedures, and the need for future AI detection designs. The scenarios provided a useful scaffold for teachers to dive into deep conversation quickly because they grounded discussion in specific situations rather than abstract questions. This structure appeared to support more actionable recommendations and avoided the hesitancy that likely would have transpired if we simply asked the group, “What are your recommendations for AI policy?”
Teacher-Generated Recommendations for Policy and Practice
We found that, across both scenarios, teachers identified four common areas of need related to policy, practice, and the design of professional development, curriculum, and school-based technology supports. The following recommendations reflect teachers’ perspectives generated through the codesign process. We present them here as empirical findings. In the next section, we extend these ideas to consider implications for teacher education, practice, and policy.
AI literacy education for teachers, administrators, and community members. Similar to the movement for digital literacy education, AI literacy education could support people to be more comfortable using AI and become more critical consumers when they see potential deepfakes or other suspicious AI-generated content. AI literacy education could also help people understand how easily the average person can create believable deepfakes. Such literacy might also help mitigate the public and reputational damage that can be caused by deepfakes. For example, the teachers shared that if the community is unaware that a teenager could create a deepfakes, then the victim might not be believed and supported when they report that the image or video is not real. Teachers identified the need for free AI literacy lessons and activities, such as those created by Common Sense Media (2025).
AI ethics education for students. Teachers suggested that students may need coaching on the difference between a joke and harassment, especially when the content is created and shared online. Building on existing anticyberbullying education (e.g., Savage et al., 2017), they emphasized that this ethical training should also highlight the potential ramifications creating such material can have on a person’s life and career. As a recent teacher-facing resource noted, “Many young people may not realize that ‘just messing around’ with AI tools to generate fake nudes of classmates or others can cross legal lines, with life-altering consequences” (Lieberman, 2025, para. 10). Teachers also saw realistic scenarios like these as useful starting points for talking with students about AI ethics, consent, and the harms associated with deepfakes.
Legal protections for victims. Teachers were concerned that individuals targeted by deepfakes may have limited recourse to remove harmful images or seek redress. They emphasized that victims should have the right to delete deepfakes and regain control of their image, whether through a program like the National Center for Missing and Exploited Children’s Take it Down initiative (2024) or stronger legal protections.
They also raised concerns about the lack of legislation around GenAI and people’s rights to their own image. As they discussed, victims may still face practical barriers, such as identifying the perpetrator, navigating reporting processes, ensuring platform compliance with takedown requests, and understanding whether any legal remedy is available, even as antideepfake laws continue to be proposed and debated (e.g., Michel, 2024; Pahwa, 2025). They viewed this uncertainty as particularly troubling in cases of nonconsensual sexual deepfakes. One teacher asked why a person could not request that their image be deleted from the internet, prompting discussion of the European Union’s Right to Erasure or Right to be Forgotten policy, an idea many teachers wished were applicable in their context. Teachers also emphasized the need for clear school procedures to hold creators accountable and to repair harm through restorative responses such as mediation or public apology.
AI detection tools. Teachers wanted their school administrators to have access to AI detection tools that can help them make quick and reliable decisions in response to suspected deepfakes. They were concerned, however, that existing tools remain too costly and technically complex for widespread use among school districts (e.g., Hutson, 2023; Rana et al., 2022; Vahdati et al., 2024). Additionally, teachers expressed concern that detection techniques may quickly become outdated as GenAI models evolves and as developers find ways to evade detection.
In discussing these limitations, they recommended that AI developers focus on tools that are low cost and feasible for school use and paired with complementary supports such as AI regulation and watermarking, though we know questions remain about efficacy and enforceability of those approaches (Thakkar & Kaur, 2024; Wu, et al., 2024). Teachers also discussed partnerships among schools or districts to collectively fund detection tool subscriptions.
Implications for Technology in Teacher Education, Practice, and Policy
Building on teachers’ recommendations, we pose implications for technology teacher education, school-based practice, and policy. Through this study, we identified a need for more intentional and coordinated approaches to preparing educators for the increasing presence of AI-generated content, particularly in relation to privacy, consent, and harm. While teachers in our study emphasized the need for AI literacy education, ethical guidance, legal protections, and detection tools, these recommendations also expose broader gaps in teacher preparation, school-based practice, and policy.
Implications for Technology Teacher Education
Our findings highlight a clear need for teacher education on AI literacy and ethics that addresses not only technical knowledge and skills but also the emotional dimensions of encountering technology-facilitated privacy harms. AI should be positioned as a sociotechnical issue that intersects with ethics, privacy, student wellbeing, and teachers’ own online lives. Teachers’ reactions in this pilot study suggest that teacher education should create space for processing vulnerability, understanding privacy violations, and building a sense of agency when responding to deepfake-related incidents.
Consistent with prior work on teacher learning and self-efficacy (e.g., Byrne et al., 2025; Tschannen-Moran & McMaster, 2009), our findings suggest that collaborative, scenario-based approaches support teachers’ sensemaking and confidence in engaging with unfamiliar technologies. This approach also shares features with case-based and problem-based learning, in which educators work through realistic and often ill-structured situations in order to develop professional judgment and apply knowledge in context (e.g., Hmelo et al., 2000; Hmelo-Silver et al., 2007).
We encourage teacher education programs to adopt scenario-based models like the one used in our study, as these realistic cases empowered teachers to engage deeply with ethical questions, anticipate school-based challenges, and generate actionable responses. Such approaches can support preservice teachers in developing both professional judgment and practical strategies for navigating uncertain technologies.
Implications for Practice (Schools and Districts)
Our findings highlight the need for proactive, coordinated, and clear procedures for responding to suspected deepfakes, including guidance on verification processes, communication with families, and coordination across roles (e.g., administrators, counselors, and school resource officers). Teachers in this study were unsure how to verify suspected deepfakes, who should be involved in responding, and how to manage incidents that originate outside of school but affect the school community.
As illustrated in the codesigned scenarios, delays in verification or unclear procedures can exacerbate harm, particularly when reputational damage spreads quickly through social media networks. The teachers’ scenarios further illustrate gaps reporting pathways and coordination across administrators, as well as a lack of clarify around individuals’ rights and legal protections. Teachers’ concerns underscore the urgency of proactive policies that help schools support students affected by nonconsensual synthetic media and advise families when incidents arise.
Implications for Policy
Beyond school-level procedures, teachers’ concerns underscore the need for clearer protections related to the misuse of individuals’ likeness in AI-generated media. While AI-related legislation is nascent (e.g., Michel, 2024), participants were generally unaware of existing protections and available recourse. These findings demonstrate the need for policy that is both more enforceable and usable in educational contexts, including state and federal policy that gives individuals meaningful control their likeness such as takedown processes, reporting mechanisms, and redress when their image is manipulated or distributed without consent. Without these protections, educators and students remain vulnerable to ongoing harm, even after incidents are identified.
Limitations
Our study also reveals important limitations and opportunities for future research. The teachers in our codesign session were not representative of a general teacher population. The sample participation (N = 9 teachers) was small and is, therefore, not generalizable. The teachers who participated were likely more AI literate due to their concurrent participation in another AI-focused program. Future researchers should investigate teachers’ awareness of and responses to deepfakes in more typical school contexts. In particular, studies involving teachers who share the same institutional environment may provide more situated insights into how school culture, resources, codes of conduct, and community politics shape responses to deepfake incidents. Such work could clarify which teachers or administrators would be contacted during a deepfake incident and how the school would realistically respond within the sociopolitical context of their community. Future research should also investigate larger samples of teachers in a variety of contexts and settings.
Conclusion
We presented qualitative findings from a codesign study with nine high school teachers from public school districts in the Mid-Atlantic US. Through discussion and scenario design, participating teachers engaged made sense of AI-generated synthetic media and deepfakes as an emerging form of interpersonal harm in high school contexts. Our findings demonstrate that teachers were able to quickly build a working understanding of deepfakes, identify key concerns related to privacy and harm, and generate recommendations for AI education, AI-related policy, and AI detection procedures and tools. We concluded by posing the recommendations for technology teacher education and professional development.
References
Alexander, S. (2025). Deepfake cyberbullying: The psychological toll on students and institutional challenges of AI-driven harassment. The Clearing House: A Journal of Educational Strategies, Issues and Ideas, 98(2), 36-50. https://doi.org/10.1080/00098655.2025.2488777
Ali, S., DiPaola, D., Lee, I., Sindato, V., Kim, G., Blumofe, R., & Breazeal, C. (2021). Children as creators, thinkers and citizens in an AI-driven future. Computers and Education: Artificial Intelligence, 2, 1-11. https://doi.org/10.1016/j.caeai.2021.100040
Alipan, A., Skues, J. L., & Theiler, S. (2018). “They will find another way to hurt you”: Emerging adults’ perceptions of coping with cyberbullying. Emerging Adulthood, 1-13. https://doi.org/10.1177/2167696818816896
Banerji, O., (2025, March 3). More teens than you think have been ‘deepfake’ targets. Education Week. www.edweek.org/technology/more-teens-than-you-think-have-been-deepfake-targets/2025/03?utm_source=chatgpt.com
Bauman, S., & Baldasare, A. (2015). Cyber aggression among college students: Demographic differences, predictors of distress, and the role of the university. Journal of College Student Development, 56(4), 317-330. https://doi.org/10.1353/csd.2015.0039
Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77-101. https://doi.org/10.1191/1478088706qp063oa
Burgess, M. (2026, April 15). The deepfake nudes crisis in schools is much worse than you thought. Wired. https://www.wired.com/story/deepfake-nudify-schools-global-crisis/
Byers, D. S., & Cerulli, M. (2021). Staying in their own lane: Ethical reasoning among college students witnessing cyberbullying. Journal of Diversity in Higher Education, 14(4), 508. https://doi.org/10.1037/dhe0000180
Byrne, V. L. (2020). Blocking and self-silencing: Undergraduate students’ cyberbullying victimization and coping strategies. TechTrends, 65(2), 164-173. doi.org/10.1007/s11528-020-00560-x
Byrne, V. L., & Hollingsworth, J. (2025). Actionable insights regarding cyberbullying among college students: A systematic review of the literature. Violence & Victims, 40(3), 477–500. doi.org/10.1891/VV-2024-0136
Byrne, V. L., Hollingsworth, J., & Kumar, P. C. (2023). Navigating tensions between protecting students from online harassment and respecting their privacy. British Journal of Technology, 1-17. doi.org/10.1111/bjet.13377
Byrne, V. L., Wiseman, D. L., Ketelhut, D. K., Moncrieffe, K., & Randolph, B. (2025). Supporting K12 teachers’ remote and hybrid teaching self-efficacy with an online professional development program.Journal on Online Learning Research, 11(1), 81-104. www.learntechlib.org/primary/p/225059
Chiu, T. K. F., Ahmad, Z., Ismailov, M., & Sanusi, I. T. (2024). What are artificial intelligence literacy and competency? A comprehensive framework to support them. Computers and Education Open, 6, 100171. doi.org/10.1016/j.caeo.2024.100171
Common Sense Media. (2025, May 25). AI literacy lessons for grades 6–12. https://www.commonsense.org/education/collections/ai-literacy-lessons-for-grades-6-12
Cover, R. (2022). Deepfake culture: The emergence of audio-video deception as an object of social anxiety and regulation. Continuum, 36(4), 609–621. https://doi.org/10.1080/10304312.2022.2084039
Dilek, M., Baran, E., & Aleman, E. (2025). AI literacy in teacher education: Empowering educators through critical co-discovery. Journal of Teacher Education, 76(3), 294–311. https://doi.org/10.1177/00224871251325083
Diaz, J. (2024, April 26). A Baltimore-area teacher is accused of using AI to make his boss appear racist. National Public Radio (NPR). https://www.npr.org/2024/04/26/1247237175/baltimore-ai-generated-racist-audio-crime
Elliott, V. (2024). Celebrity deepfake porn cases will be investigated by meta oversight board. Wired. https://www.wired.com/story/meta-oversight-board-deepfake-porn-facebook-instagram/
Foley, J. (2024, April 18). 23 of the best deepfake examples that terrified and amused the internet. Creative Bloq. https://www.creativebloq.com/features/deepfake-examples
Freire, P., Macedo, D.P., & Shor, I. (2018), Pedagogy of the oppressed (trans. By M. B. Ramos; 50th anniversary ed.). Bloomsbury Academic.
Gee, J. P. (2015). The new literacy studies 1. In [EDITOR NAME(S)] The Routledge handbook of literacy studies (pp. 35-48). Routledge.
Gibson, C., Olszewski, D., Brigham, N. G., Crowder, A., Butler, K. R., Traynor, P., [PER APA LIST ALL AUTHORS UP TO 19]… & Kohno, T. (2024). Analyzing the AI nudification application ecosystem. arXiv preprint arXiv:2411.09751. https://doi.org/10.48550/arXiv.2411.09751
Glynn, P. (2026, March 17). ‘Babe, they’re AI’ – Zendaya says it wasn’t just fans fooled by fake wedding photos. BBC. https://www.bbc.com/news/articles/clyz7llp4k2o
Guha, M. L., Druin, A., & Fails, J. A. (2013). Cooperative inquiry revisited: Reflections of the past and guidelines for the future of intergenerational co-design. International Journal of Child-Computer Interaction, 1, 14–23. https://doi.org/10.1016/j.ijcci.2012.08.003
Hmelo, C., Holton, D.L., & Kolodner, J. (2000). Designing to learn about complex systems. Journal of the Learning Sciences, 9(3), 247-298. https://doi.org/10.1207/S15327809JLS0903_2
Hmelo-Silver, C. E., Duncan, R. G., & Chinn, C. A. (2007). Scaffolding and achievement in problem-based and inquiry learning: A response to Kirschner, Sweller, and Clark. Educational Psychologist, 42(2), 99–107. https://doi.org/10.1080/00461520701263368
Holt, M. K., Greif Green, J., Reid, G., DiMeo, A., Espelage, D. L., Felix, E. D., Furlong, M. J., Poteat, V. P., & Sharkey, J. D. (2014). Associations between past bullying experiences and psychosocial and academic functioning among college students. Journal of American College Health, 62(8), 552–560. https://doi.org/10.1080/07448481.2014.94799
Hutson, M. (2023, March 6). Detection stays one step ahead of deepfakes—For now. IEEE spectrum. https://spectrum.ieee.org/deepfake
Hwang, Y., Ryu, J. Y., & Jeong, S.-H. (2021). Effects of disinformation using deepfake: The protective effect of media literacy education. Cyberpsychology, Behavior, and Social Networking, 24(3), 188–193. https://doi.org/10.1089/cyber.2020.0174
Kennedy, K. J. & Castek, J. M. (2025).Empowering ELA teachers: Recommendations for teacher education in the AI era. Contemporary Issues in Technology and Teacher Education, 25(1), 1-31. https://citejournal.org/volume-25/issue-1-25/english-language-arts/empowering-ela-teachers-recommendations-for-teacher-education-in-the-ai-era/
Knight, W. (2018, August). Fake America great again. MIT Technology Review. www.technologyreview.com/2018/08/17/240305/fake-america-great-again
Kumar, P., & Byrne, V. L.(2022). The 5Ds of privacy literacy: A framework for privacy education. Information and Learning Sciences. doi.org/10.1108/ILS-02-2022-0022
Laupichler, M. C., Aster, A., Schirch, J., & Raupach, T. (2022). Artificial intelligence literacy in higher and adult education: A scoping literature review. Computers and Education: Artificial Intelligence, 3, 100101. doi.org/10.1016/j.caeai.2022.100101
Le, T. H., Huynh, L., Dang, B., Pham, H.-H., Nguyen, N. T. V., & Nguyen, A. (n.d.). Development and evaluation of artificial intelligence literacy training for teacher education students. British Journal of Educational Technology (early access).https://doi.org/10.1111/bjet.70047
Lieberman, B. (2025, May 15). Deepfakes can be a crime – Teaching AI literacy can prevent it. Common Sense Media. www.commonsensemedia.org/kids-action/articles/deepfakes-can-be-a-crime-teaching-ai-literacy-can-prevent-it
Long, D., Teachey, A., & Magerko, B. (2022, April). Family learning talk in AI literacy learning activities. In [EDITOR NAME(S)?] Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems (pp. 1-20). https://doi.org/10.1145/3491102.3502091
Marder, B., Joinson, A., Shankar, A., & Houghton, D. (2016). The extended “chilling” effect of Facebook: The cold reality of ubiquitous social networking. Computers in Human Behavior, 60, 582–592. https://doi.org/10.1016/j.chb.2016.02.097
Michel, A. H. (2024, July 24). The ACLU fights for your constitutional right to make deepfakes. Wired. https://www.wired.com/story/aclu-artificial-intelligence-deepfakes-free-speech/
Mills, K., Ruiz, P., Lee, K. W., Coenraad, M., Fusco, J., Roschelle, J., & Weisgrau, J. (2024). AI literacy: A framework to understand, evaluate, and use emerging technology. Digital Promise. https://doi.org/10.51388/20.500.12265/218
National Center for Missing & Exploited Children. (2024). Take it down. https://takeitdown.ncmec.org/
Nissenbaum, H. (2010). Privacy in context: Technology, policy, and the integrity of social life. Stanford University Press.
Ng, D. T. K., Leung, J. K. L., Chu, K. W. S., & Qiao, M. S. (2021a). AI literacy: Definition, teaching, evaluation and ethical issues. Proceedings of the Association for Information Science and Technology, 58(1), 504–509. https://doi.org/10.1002/pra2.487
Ng, D. T. K., Leung, J. K. L., Chu, S. K. W., & Qiao, M. S. (2021b). Conceptualizing AI literacy: An exploratory review. Computers and Education: Artificial Intelligence, 2, 100041. https://doi.org/10.1016/j.caeai.2021.100041
Pahwa, N. (2025, May 22). Congress passed a sweeping free-speech crackdown—and no one’s talking about it. Slate. https://slate.com/technology/2025/05/deepfake-trump-take-it-down-act-revenge-porn-explained.html
Pandey, C. K., Mishra, V. K., & Tiwari, N. K. (2021). Deepfakes: When to use it. 2021 10th International Conference on System Modeling & Advancement in Research Trends (SMART), 80–84. doi.org/10.1109/SMART52563.2021.9676297
Penuel, W. R., Roschelle, J., & Shechtman, N. (2007). Designing formative assessment software with teachers: An analysis of the co-design process. Research and Practice in Technology Enhanced Learning, 02(01), 51–74. https://doi.org/10.1142/S1793206807000300
Rana, M. S., Nobi, M. N., Murali, B., & Sung, A. H. (2022). Deepfake detection: A systematic literature review. IEEE Access, 10, 25494–25513. IEEE Access. doi.org/10.1109/ACCESS.2022.3154404
Ray, K. (2023, August 22). Is generative AI the most-wanted cyberbully? Tech & Learning. https://www.techlearning.com/news/is-generative-ai-the-most-wanted-cyberbully
Reface. (2025). https://reface.ai/
Reiser, B., & Tabak, I. (2014). Scaffolding. In R. K. Sawyer (Ed.), The Cambridge handbook of the learning sciences (2nd ed; pp. 44–62). Cambridge University Press.
Saldaña, J. (2013). The coding manual for qualitative researchers (2nd ed). SAGE
Savage, M. W., Deiss Jr, D. M., Roberto, A. J., & Aboujaoude, E. (2017). Theory-based formative research on an anti-cyberbullying victimization intervention message. Journal of Health Communication, 22(2), 124-134. https://doi.org/10.1080/10810730.2016.1252818
Scolforo, M. (2026, March 26). Pennsylvania teens get probation after using AI to create fake nudes of classmates. Associated Press. https://6abc.com/post/lancaster-pennsylvania-teens-get-probation-using-ai-create-fake-nudes-classmates/18778852/
Singer, N. (2024a, April 8). Teen girls confront an epidemic of deepfake nudes in schools. The New York Times. https://www.nytimes.com/2024/04/08/technology/deepfake-ai-nudes-westfield-high-school.html
Singer, N. (2024b, April 22). Spurred by teen girls, states move to ban deepfake nudes. The New York Times. https://www.nytimes.com/2024/04/22/technology/deepfake-ai-nudes-high-school-laws.html
Smith, J., & Yoon, J. (2013). Cyberbullying presence, extent, and forms in a Midwestern post-secondary institution. Information Systems Education Journal, 11(3), 52. https://files.eric.ed.gov/fulltext/EJ1145015.pdf
Tagare, D., Karki, T., & Yu, W. (2025). K-12 teachers’ ethical competencies for AI literacy: Insights from a systematic literature review. Computers & Education, 239, 105435. https://doi.org/10.1016/j.compedu.2025.105435
Thakkar, J. J., & Kaur, A. (2024). From deepfakes to digital truths: The role of watermarking in AI-generated image verification. In Proceeding of the 2024 47th International Conference on Telecommunications and Signal Processing (pp. 216–222). doi.org/10.1109/TSP63128.2024.10605975
Thorn. (2025). Deepfake nudes & young people: Navigating a new frontier in technology-facilitated nonconsensual sexual abuse and exploitation. https://www.thorn.org/research/library/deepfake-nudes-and-young-people/
Trammel, J. M. (2023). Artificial intelligence for social evil: exploring how AI and beauty filters perpetuate colorism—lessons learned from a colorism giant, Brazil. In K. Langmia (Ed.), Black communication in the age of disinformation: Deepfakes and synthetic media (pp. 51-71). Springer International Publishing.
Tschannen‐Moran, M., & McMaster, P. (2009). Sources of self‐efficacy: Four professional development formats and their relationship to self‐efficacy and implementation of a new teaching strategy. The Elementary School Journal, 110(2), 228–245. doi.org/10.1086/605771
U.S. Department of Education, Office of Educational Technology. (2024). Empowering education leaders: A toolkit for safe, ethical, and equitable AI integration. https://files.eric.ed.gov/fulltext/ED661924.pdf
Vahdati, D. S., Nguyen, T. D., Azizpour, A., & Stamm, M. C. (2024). Beyond deepfake images: Detecting AI-generated videos. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 4397-4408). doi.org/10.48550/arXiv.2404.15955
Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Harvard University Press. https://doi.org/10.1007/978-3-540-92784-6
Wang, B., Rau, P.-L. P., & Yuan, T. (2023). Measuring user competence in using artificial intelligence: Validity and reliability of artificial intelligence literacy scale. Behaviour & Information Technology, 42(9), 1324–1337. doi.org/10.1080/0144929X.2022.2072768
Wei, M., Yeung, C., Roesner, F., & Kohno, T. (2025, April). “We’re utterly ill-prepared to deal with something like this”: Teachers’ perspectives on student generation of synthetic nonconsensual explicit imagery. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (pp. 1-18). doi.org/10.1145/3706598.3713226
Wilson, S. & Doran, B. (2024, November 13). Lancaster county parents call for change after deepfake pornographic images of over 50 female students created. ABC27. https://www.abc27.com/local-news/lancaster/lancaster-county-parents-call-for-change-after-deepfake-pornographic-images-of-over-50-female-students-created/
Wineburg, S., & McGrew, S. (2019). Lateral reading and the nature of expertise: Reading less and learning more when evaluating digital information. Teachers College Record, 121(11), 1-40. https://doi.org/10.1177/016146811912101102
Wu, X., Liao, X., Ou, B., Liu, Y., & Qin, Z. (2024). Are watermarks bugs for deepfake detectors? Rethinking proactive forensics (arXiv:2404.17867). arXiv. https://doi.org/10.48550/arXiv.2404.17867
