除了发现深度伪造：教导多语言学习者质疑人工智能
原文标题: Beyond spotting deepfakes: Teaching multilingual learners to interrogate AI
来源: eSchoolNews | 发布时间: 2026-08-25
原文链接: 点击阅读原文
Share on X (Opens in new window)X
Share on Facebook (Opens in new window)Facebook
Share on LinkedIn (Opens in new window)LinkedIn
Email a link to a friend (Opens in new window)Email
Print (Opens in new window)Print
Key points:
When it comes to AI, visual access must be paired with visual questioning
5 ways AI can strengthen your teaching this school year
Schools are building AI rules before they know the destination
For more on AI in the classroom, visit eSN’sDigital Learninghub
Long before computers entered our homes and classrooms, photographs helped shape what people believed. Images carried a particular kind of authority: We did not simply look at them; we often treated them as evidence. They helped audiences understand events they had never witnessed themselves and gave distant stories a sense of immediacy and truth.
Yet images have never been neutral witnesses. They can be selected, cropped, staged, retouched, captioned, or removed from their original context. Photographic manipulation predates the digital age. What artificial intelligence has changed is not the possibility of visual deception, but its speed, accessibility, scale, and increasingly convincing realism.
Today, students can encounter an AI-generated photograph, a cloned voice, or a convincing video of an event that never occurred—all while scrolling through the same feed where they see authentic news and messages from friends.
For multilingual learners, this deserves particular attention. Visuals often bridge linguistic gaps, making concepts accessible before students have the language to fully explain them.
But when images can be generated, manipulated, or taken out of context, visual access must be paired with visual questioning.
This is not because multilingual learners are inherently more susceptible to misinformation.
Rather, if we use visuals as a bridge to understanding, we must also teach students to question their source, context, and credibility.
From looking to investigating
One response to deepfakes is to teach students how to spot them: Look for distorted hands, unusual shadows, unnatural facial movements, or inconsistencies in the background.
These clues may help, but they have a limitation: AI is improving. The visual flaw that exposes an AI-generated image today may disappear tomorrow.
Students therefore need to move beyond asking:
“Does this look real?” and learn to ask: “What evidence do I have that this is real?”
Research from Stanford’sCivic Online Reasoninginitiative found that students often struggled to judge the credibility of online information, challenging the assumption that growing up with technology makes young people skilled evaluators. Professionalfact-checkers used a different strategy: lateral reading—leaving a questionable page to investigate its source and seek independent evidence.
Applied to deepfakes, the lesson is simple: Don’t just look more closely at the image.
Look beyond it.
For multilingual learners, teachers can make both the thinking and language of verification explicit: What do I see? What is the claim? Who created it? Where else can I check? What evidence do I have? Students can discuss their reasoning with a partner or in their strongest language before communicating conclusions in English, supported by frames such as This source claims…, I questioned this because…, and The evidence does/does not support…
Google Arts & Culture’s Odd One Out, which asks students to identify an AI generated image among authentic artworks, offers a low-language-entry way to practise this reasoning.
Teachers can move beyond simply spotting the “fake” by asking students what made them question the image, what evidence supports their judgment, and where else they could check.
Interpret, generate, evaluate
Questioning AI-generated content is only part of AI literacy. Students also need opportunities, with appropriate safeguards, to understand how it is constructed. A useful progression is to interpret, generate, and evaluate.
Interpret: Students examine an image or video, considering its source, framing, context, purpose, and audience. For multilingual learners, visuals provide an accessible entry point, while structured discussion develops the language needed to express increasingly complex interpretations.
Generate: Students might create two AI-generated images of the same fictional event, changing the perspective, wording, or framing of the prompt. They then compare what changed, what was emphasised or omitted, and which version appears more convincing.
The goal is not simply to learn prompting, but to experience how choices shape the story an image appears to tell.
Evaluate: Students return to the evidence: Who created this? What supports the claim? What do reliable sources say? What might be missing? What would make me reconsider my conclusion? In doing so, they use academic language to question, corroborate, challenge, justify, and revise.
This progression reflects higher-order processes in Bloom’s Revised Taxonomy: students analyse how visual meaning is constructed, apply their understanding through creation, and evaluate claims using evidence, context, and credibility.
Linguistic scaffolding does not lower expectations; it gives multilingual learners greater access to these sophisticated cognitive processes.
Beyond detecting deepfakes
These skills extend beyond AI-generated content. A genuine photograph paired with a false caption can mislead. An authentic video removed from its original context can distort an event. The larger goal is therefore not simply to teach multilingual learners how to detect AI. It is to give them the tools and language to evaluate claims and evidence.
Stanford’s research offers an encouraging message: these skills can be taught. Studies evaluating Civic Online Reasoning instruction have shown that students can improve their ability to judge the credibility of online information.
Importantly, we do not want students to leave our classrooms believing that nothing can be trusted. Skepticism without the skills to investigate can quickly become cynicism.
The goal is not: “Everything could be fake.”
It is: “I don’t know whether I should trust this yet. Let me find out.”
Protection through preparation
Schools have legitimate reasons to establish boundaries around AI, particularly around privacy, safeguarding, age, and assessment. But boundaries need to be accompanied by education.
This is especially important for multilingual learners. We have long recognised the power of visuals to make learning more accessible. In the age of generative AI, we need to extend that principle: access to visual information must be accompanied by the tools to interrogate it.
Keeping AI outside the classroom will not keep it outside students’ lives.
We do students a disservice if they leave school knowing how to avoid AI but not how to question an AI-generated image, investigate its source, corroborate its claims, or decide whether it should be shared.
Our responsibility is to give every learner—including those still developing the language of instruction—the tools and language to interpret what they see, generate responsibly, investigate what they encounter, and evaluate what deserves their trust.
Because in the age of deepfakes, perhaps the most important question we can teach a student to ask is not: “Is this real?” but: “How do I know?”
Sign up for our K-12 newsletter
Δ
Nesreen El-Baz is an ESL educator with over 20 years of experience, and is a certified bilingual teacher with a Master’s in Curriculum and Instruction. El-Baz is currently based in the UK, holds a Masters degree in Curriculum and Instruction from Houston Christian University, and specializes in developing in innovative strategies for English Learners and Bilingual education.
Students won’t remember every lesson, but here’s what’ll stick- August 27, 2026
Building inclusive school communities through integrated digital strategy- August 26, 2026
Beyond spotting deepfakes: Teaching multilingual learners to interrogate AI- August 25, 2026
Share on X (Opens in new window)X
Share on Facebook (Opens in new window)Facebook
Share on LinkedIn (Opens in new window)LinkedIn
Email a link to a friend (Opens in new window)Email
Print (Opens in new window)Print
Want to share a great resource? Let us know atsubmissions@eschoolmedia.com.
本报道由 AI 助手自动抓取、翻译并发布。