📮 Homework · Deepfakes: pictures, videos and voices made by AI

You will ask three adults at home the question, 'what is a deepfake?' Write down each answer word for word. Bring the three answers back. Compare them with the class definition. Note which adult's answer was closest, and which was furthest, and one detail every adult missed.

You will ask one adult at home how they would tell whether a video of a well-known person is real. Write down their answer word for word. Then, without correcting them, gently ask what they would do if their first check failed. Bring both answers back. Compare with the three habits in class. Note one habit the adult mentioned and any they missed.

You will find one other publicly reported deepfake case from the last two years using a school-safe search on BBC News, the Guardian, or the National Cyber Security Centre website. Write down: who was faked, how the fake landed, what harm followed, and what verification would have caught it. Bring the write-up back for a class comparison. Do not share links or details from personal social media.

You will do this with a grown-up at home. Together, you will draw your own pretend picture of a pet or a toy. It does not have to be tidy. When it is finished, your grown-up helps you write the word PRETEND under it. Bring it back to school so Miss can put it on the wall.

You will do this at home with a grown-up. Draw your pet, or a made-up pet, twice. Once as a cartoon with big eyes. Once as best you can as a real-looking drawing. Show both to a grown-up. Ask which one looks more real. Bring both back to school and write PRETEND on the back of each.

You will explain the word deepfake to one adult at home in two sentences of your own. Then ask them if they have ever seen one. Write down what they say. Bring the answer back and share it in class. If the adult has not seen one, that is fine, just note that.

You will ask three adults at home the question, 'if you close your eyes and picture a scientist at work, what do you see?' Write down each description. Bring them back to compare with the AI's scientist and with the scientist you drew. Note which adult described someone the AI did not draw.

You will do this at home with a grown-up. Ask the grown-up to run the prompt 'a family photograph in the living room at Christmas' three times using any AI image tool your family already has. Look at the three pictures together. Tally: number of adults, number of children, ages, setting. Write one sentence about who the AI kept drawing and who it missed.

You will take three pictures at home: one real photograph, one drawing or painting, and one printed AI image your teacher gives you. Underneath each, write the honest four-part caption. Bring all three captioned pictures back for peer marking in class. If you have no printer, describe the pictures in words on paper and caption the descriptions.

You will ask two adults at home how they would tell whether a photograph of a well-known person is real or AI-generated. Write down their answers. Do they mention the 'blurry eyes' or 'wonky hands' rule? Bring the answers back. In class, we will discuss which of their rules have quietly expired and which habits still work.

You will find one place a young person can go for help if a fake AI image of them is circulating. Write down: the name, what they do, and how to contact them. Two safe suggestions: the school's Designated Safeguarding Lead, and Childline on 0800 1111 (or childline.org.uk). Bring the note back for a class safeguarding wall. Report Harmful Content (reportharmfulcontent.com) is another safe route to know.

You will find one short primer on diffusion models from a trusted source (a BBC News technology explainer, a Royal Society summary, a school-approved Computing resource). Read or watch it once. Write three things it explained that surprised you. Bring the notes to class for a comparison. Do not search random YouTube videos for this task.

You will have a conversation at home with an adult about deepfakes of people at school. Agree together what your family's plan is if one appears in a group chat you are in. Write down the plan in three lines. Bring it back for class comparison. If your adult has not thought about this before, that is normal; the conversation IS the point.

You will take three news photographs from a school-safe UK news site (BBC, Guardian, ITV) and reverse-image-search each on the school device before you leave school, or on a family device with adult permission. Record: matches found, sources, and any inconsistencies. Bring the results back for class comparison. Do not run this on private social media photos of real people.

You will ask one adult at home whether they have ever agreed for their photograph to be used to train an AI. Most will say no. Then ask whether they have ever posted a photograph publicly online. Most will say yes. Write both answers, and one sentence about the gap between them. Bring it back for class discussion.

You will pick one striking image from a UK news homepage tonight (BBC, Guardian, ITV). Run the four-step chain on it. Log each step. Bring the log back. Report whether the image passed all four. Do not run this on private social media photos of ordinary people, only on published news images.

You will pick one of three UK safeguarding routes and learn it well enough to explain in class: (1) Report Remove for under-eighteens at childline.org.uk/report-remove, (2) Report Harmful Content at reportharmfulcontent.com, (3) Childline on 0800 1111 or childline.org.uk. Write down: what they do, when to use them, and how to contact them. Bring it back to teach the class.

You will write, at home, a five-mark model answer to: 'Explain how a modern deepfake video is generated from a small amount of source material. Refer to the components of the pipeline and the concept of a latent representation.' Include a diagram. Bring it in for peer marking. Aim for GCSE Computer Science register.

You will read the current AI-content or synthetic-media policies of two major platforms (TikTok, Meta, X, YouTube) using their published policy pages. Note: what they promise, what they enforce automatically, and what a user has to do. Bring a one-page comparison back for the class discussion. Use only the platforms' own policy pages.

You will research one publicly reported election-period deepfake from a UK, US or European election in the last three years, using BBC News or the Guardian. Note: what the fake claimed, how fast it spread, how fast it was debunked, and whether it appears to have affected the result. Bring the note back for class discussion.

You will find one clearly satirical UK piece (a Spitting Image episode, a Have I Got News For You segment, a BBC political sketch) and one paragraph in a school-safe news write-up of a harmful deepfake incident. Write a 200-word register analysis comparing the two. Bring it in for peer marking. Do not include the actual harmful content, only the news write-up description.

You will have the safe-word conversation at home tonight with the adults you live with. Agree a word, agree the call-back habit, and agree what happens if a voicenote arrives without the word. Do NOT bring the word back to class, but bring a one-line confirmation that the conversation happened. This is the highest-impact ten-minute conversation on this topic your family will have.

You will write up your experiment as a two-page A-level Computer Science report: hypothesis, method, results, discussion, and one paragraph on the professional implication (why provenance beats detection). Include the two images and their scores. Bring for peer marking. Do not include any faces of real people in your prompts.

You will visit Reuters Pictures, Getty News, or PA Media online and browse one real news-wire event feed with at least a dozen images. Note the compositional variety in the real feed. Compare with the six-image AI gallery in class. Write a one-page comparative reflection. Bring it back for peer discussion.

You will read Walter Benjamin's 'The Work of Art in the Age of Mechanical Reproduction' (any section) plus one recent (post-2023) essay on generative AI and authenticity from a school-approved philosophy source. Write a 400-word reflection connecting the two. Bring in for peer discussion.

You will find one published Ofcom decision or enforcement notice from the last twelve months and summarise it in 400 words: the platform, the content, the OSA provisions engaged, and the outcome. Bring the summary in for peer marking. Use only Ofcom's published enforcement pages.

You will read one primary source paper on one of the four mechanisms (Hasher, Goldstein & Toppino 1977 on illusory-truth is a good starting point). Write a 400-word summary linking the mechanism to a current deepfake case. Bring for peer marking. Use only school-approved academic sources.

You will introduce your chosen habit at your workplace this week (or at your placement if you are a learner). Report back at the next session on: what happened, how colleagues responded, and one refinement. This is a live workforce exercise; the report is confidential to the class.

You will locate the MHRA public alerts feed at gov.uk/drug-safety-update and bookmark it on your work-approved device. Note: how often it updates, how to search by drug name, and how to sign up for email alerts. Bring the reference back for the team log. This becomes part of your standing verification kit.

You will audit your actual site (or placement) against the standing order you drafted in class. Note: which controls are in place, which are informal, and which are missing. Bring the audit back for peer discussion. Anonymise any site or supplier details.

You will draft, at home, the one-page defeat pipeline for your own firm (or placement). Include all four steps with the specific controls your firm has today, plus a note on where the gaps are. Bring for team discussion. Anonymise any specific supplier or customer details.

You will pick one platform your organisation uses (LinkedIn, X, an internal SharePoint, a CMS) and test empirically whether it preserves C2PA metadata on upload. Method: create a small test file with content credentials, upload, download the served version, run through contentcredentials.org/verify. Log the result. Bring back for peer discussion and add to your K13 evidence.

You will draft, at home, the four-stage detection pipeline as a one-page procedure for your own team's next fine-tuning task. Include: the rule set, the detector, the stylometric features, the human-sampling protocol, and the contamination risk score threshold. Bring for peer critique and EPA-portfolio consideration.

You will audit your own team's actual controls at the layer you drafted evidence for in class. Note: which controls exist, which are documented, which are informal. Bring for peer discussion. Anonymise team-member names and specific system identifiers.