Module 2 · AI in Our World — Basic | AESOP AI Academy Module 2
Color
Basic
Lesson 1
·Quiz·Lab
Lesson 2
·Quiz·Lab
Lesson 3
·Quiz·Lab
Lesson 4
·Quiz·Lab
Lesson 5
·Quiz·Lab
Lesson 6
·Quiz·Lab
Module Test
🔬 Basic
Lesson 1: AI You See Every Day
AI isn't just robots and science fiction — it's quietly running in almost every app and device you use right now.
Keisha has the same phone as her cousin Jaylen, and they both use the same music app. But their "Recommended for You" lists look completely different. Jaylen gets hip-hop and R&B. Keisha gets pop and video game soundtracks. Same app. Same algorithm. Completely different results.
At first Keisha thought it was just random. Then she realized: the app is learning. Every song she skips, every playlist she finishes, every time she hits replay — it's all data. The app is building a model of what Keisha likes. And it's doing that for every single user, all at the same time.
AI Is Already Everywhere
You probably interact with AI dozens of times before breakfast. Your phone's autocorrect uses a language model to predict your next word. Your email spam filter uses a classifier trained on billions of emails to decide what's junk. Your streaming app uses a recommendation system to decide what to show you. Your social media feed uses an algorithm to rank what you see first.
None of these announce themselves as AI. They just work — quietly, continuously, shaping what you see and read and hear. This invisible layer is sometimes called ambient AI: AI so woven into the infrastructure that it's just the way things work.
🔑 What Is a Recommendation Algorithm?
The most common type is called collaborative filtering. It doesn't try to understand whether content is good — it looks at your past behavior, finds other users who behaved similarly, and recommends what they liked. It's pattern-matching on a massive scale. The result: the app seems to "know" you, even though it's just doing statistics.
What Is the Algorithm Actually Optimizing For?
Here's the important thing to understand: recommendation algorithms aren't trying to show you the best content. They're trying to show you content that keeps you on the app as long as possible — because that's how the company makes money. Usually, those two goals overlap. But not always.
Research has found that some platforms' algorithms amplify content that makes people angry or upset, because strong emotions cause people to keep scrolling. The algorithm isn't "evil" — it's just doing what it was designed to do: maximize engagement. But what it's designed to do isn't the same as what's good for you.
🌟 Think About It
After learning this, how do you feel about scrolling through your feed? Knowing an algorithm is deciding what you see, designed to keep you watching — does that change anything about how you want to use it?
🧪 Why did Keisha and Jaylen get different music recommendations from the same app?
✓ Correct — ✅ The algorithm tracks every user separately — each person gets recommendations based on their own history and similar users' behavior.
The algorithm tracks each user's behavior separately and builds a model of their preferences — that's why two people using the same app see different things.
🧪 What is collaborative filtering?
✓ Correct — ✅ Collaborative filtering matches you to similar users and recommends what those users engaged with — it's statistics, not taste judgment.
Collaborative filtering finds people who behave like you and recommends what they liked — it doesn't judge content quality, just patterns of engagement.
🧪 What does "ambient AI" mean?
✓ Correct — ✅ Ambient AI is everywhere in our apps and devices — autocorrect, spam filters, recommendations — operating invisibly as part of the normal experience.
Ambient AI refers to AI embedded so invisibly in everyday technology that people interact with it constantly without thinking of it as AI.
🧪 What is a recommendation algorithm actually designed to maximize?
✓ Correct — ✅ Most recommendation algorithms optimize for engagement (time on app) because that drives advertising revenue — not for what's best for users.
Recommendation algorithms are designed to maximize engagement (time on app) because that's how the company makes money — not to maximize what's best for you.
🧪 Why might a recommendation algorithm show content that makes you angry?
✓ Correct — ✅ Strong emotions, including anger, drive high engagement. An algorithm optimizing for time on app will amplify emotionally activating content even if it's upsetting.
Strong emotions keep people scrolling. An algorithm maximizing engagement will surface content that provokes strong reactions — even negative ones.
Find the hidden AI systems in your daily digital life.
Lab 1 — Spot the AI
You probably use more AI every day than you realize. In this lab, you'll tell the AI about the apps and devices you use, and together you'll figure out what AI is running inside them.
List 3–5 apps or devices you use regularly (phone, games, streaming, social media, etc.).
The AI will help you identify what kind of AI is running in each one.
Discuss: which one surprised you the most? Which one do you feel differently about now?
You don't need to share anything private — just the names of apps or types of devices you use.
🔬 Lab AssistantLab 1
Tell me 3 to 5 apps or devices you use regularly — games, streaming apps, social media, your phone features, anything. I'll help you figure out what AI is quietly running inside each one.
AI is changing how we learn and how we play — and raising some hard questions about what "help" really means.
Tomás has a book report due Friday. His friend shows him how to use an AI writing tool. He types in his prompt, reads the result, and it's good — way better organized than what he usually writes. He feels weird. Is this cheating? Is it like using spell-check, or is it something different?
He decides to try something. He submits the AI version, then goes back and writes his own version from scratch. When he reads them side by side, something interesting happens: his version is messier, yes — but it has things in it that the AI version doesn't. His actual opinion about the book. A connection to something that happened to him. The AI wrote about the book. He wrote what the book meant to him.
AI and Learning: The Core Question
Schools everywhere are figuring out how to handle AI writing tools. The honest question isn't just "is it cheating?" — it's "what is the assignment actually trying to build in you?" If the goal is to produce a polished essay, AI can help. If the goal is to practice organizing your thoughts and finding your own voice, using AI to write it defeats the purpose.
Learning scientists talk about something called desirable difficulty — the idea that struggling with a hard task, even when it's frustrating, builds stronger understanding than getting the answer handed to you. When you work through something difficult, your brain builds stronger connections. AI that makes learning frictionless can accidentally take that away.
🔑 The Difference That Matters
Using AI to skip the learning process is different from using AI to deepen it. Getting AI to write your essay skips the process. Using AI to get feedback on your draft deepens it. The question isn't whether to use AI — it's what role you're giving it in your learning.
AI in Sports and Games
AI has become a big deal in professional sports. Teams use AI to analyze players' movements, predict injuries before they happen, and study opponents' patterns. In chess and Go, AI systems have reached superhuman levels — not by copying how humans play, but by discovering completely different strategies that humans had never considered. When DeepMind's AlphaGo beat the world champion in 2016, the AI made moves that experts described as "alien" — totally unlike human play, but undeniably effective.
In video games, AI controls non-player characters (NPCs), adjusts difficulty to match your skill level, and creates procedurally generated worlds. Game AI has its own research field — and techniques developed for games have often ended up being useful in real-world AI applications.
🌟 What Does "Winning" Mean?
If an AI helps you win a game but you didn't actually get better at playing, did you win? If AI writes your assignment and you get a good grade but didn't learn anything, did you succeed at school? The tools can change the outputs without changing what they were actually supposed to build in you.
✓ Correct — ✅ Desirable difficulty: productive struggle builds stronger understanding than frictionless help — the effort is part of what creates real learning.
Desirable difficulty means the cognitive struggle of working through something hard is itself what builds strong learning — taking shortcuts can undermine this.
🧪 What did Tomás notice when he compared the AI version and his own version of the book report?
✓ Correct — ✅ The AI produced polished writing but his version contained his actual perspective — the personal meaning that AI couldn't generate.
The AI version was well organized, but Tomás's version contained his actual opinions and personal connections — what the book meant to him, which the AI couldn't know.
🧪 What's the key difference between using AI to "skip" vs. "deepen" learning?
✓ Correct — ✅ The question is whether you're using AI to replace the cognitive work or to enhance it — those lead to very different outcomes for your actual learning.
Skipping = AI does the work for you. Deepening = AI helps you do the work better. The same tool can do either depending on how you use it.
🧪 What was described as "alien" about AlphaGo's moves?
✓ Correct — ✅ AlphaGo discovered strategies that no human had ever tried — valid and brilliant, but coming from an intelligence that doesn't think the way humans do.
AlphaGo's moves followed all the rules but came from a completely different kind of problem-solving — discovering strategies humans had never thought of.
🧪 How does AI adjust difficulty in video games?
✓ Correct — ✅ Adaptive AI difficulty tracks how you're performing and adjusts the challenge to keep the game engaging — not too easy, not too hard.
AI-driven difficulty adjustment tracks your performance and modifies the challenge to match your skill level in real time.
Figure out when AI is actually helping you learn — and when it's just doing the work for you.
Lab 2 — The Learning Experiment
Think about something you're currently learning at school — a subject, a concept, a skill. Work with the AI to figure out how you could use AI tools to deepen your learning rather than skip it.
Tell the AI what you're currently learning or struggling with at school.
Together, brainstorm: how could AI help you understand it better WITHOUT just giving you the answer?
The AI will suggest some specific approaches — and ask you which ones you'd actually try.
The goal is to find ways AI can be a useful tool for your learning, not a shortcut around it.
🔬 Lab AssistantLab 2
Tell me something you're currently learning or studying at school — any subject, any grade level. I'll help you think about ways to use AI that would actually make you better at it, rather than just having AI do it for you.
Traffic systems, emergency services, local government — AI is making decisions that affect your neighborhood, often without anyone noticing.
Riley's mom calls herself a public defender. She explains cases to Riley sometimes. One day she tells Riley about a man in their city who was denied bail partly based on a computer score. The score predicted he was "high risk." Riley asks: who decided what goes into the score? Who checked if it was fair? Can the man argue against it?
Riley's mom is quiet for a second. "Those are exactly the right questions," she says. "And the answer to most of them is: not really. The company that made the software says it's proprietary — that means secret. So we can't fully check how it works. But it's affecting real people."
AI Is Making Community Decisions
AI systems are being used in communities in ways that affect people's lives significantly. In criminal courts, algorithms score how likely someone is to reoffend — and those scores can influence whether they go to jail while waiting for trial. In some cities, traffic algorithms decide which streets get priority for emergency vehicles. In social services, AI tools help decide who gets assistance first.
These applications can improve efficiency. But they raise serious questions about fairness and accountability. Who programmed the algorithm? What data was it trained on? If it makes a mistake, how does the person affected find out — or challenge it?
🔑 The Fairness Problem
A 2016 investigation found that one widely used court risk score was much more likely to incorrectly label Black defendants as "high risk" than white defendants with similar backgrounds. The company said their formula was fair. Researchers said it wasn't. Both were measuring fairness differently — and it turned out you mathematically can't satisfy both definitions at the same time when different groups have different base rates. Fairness, it turns out, isn't just a technical question.
The Accountability Gap
When a human makes a bad decision about you, there are ways to challenge it — talk to their boss, file a complaint, go to court. When an algorithm makes a bad decision, that path gets complicated. The decision was made by software. The software belongs to a private company. The company says the formula is a trade secret. So you can't see it, can't fully challenge it, and may not even know it affected you.
This gap between algorithmic decisions and human accountability is one of the most important issues in AI today. It's not about whether AI is useful in government — it often is. It's about whether the people affected by those decisions have any way to understand and contest them.
🌟 Accuracy for Whom?
A system can be accurate "on average" while still being significantly less accurate for specific groups. If an algorithm predicts crime correctly 80% of the time overall, but only 65% of the time for a specific neighborhood, those residents are getting a worse deal — even though the headline number looks fine. We have to ask not just "how accurate?" but "accurate for whom?"
🧪 What is the "accountability gap" in algorithmic decisions?
✓ Correct — ✅ The accountability gap: proprietary code in private hands makes algorithmic decisions hard to inspect, challenge, or hold accountable through normal democratic processes.
The accountability gap is the difficulty of challenging algorithmic decisions when the formula is a trade secret owned by a company — you can't see it, challenge it, or fully understand how it affected you.
🧪 What happened with the court risk score investigated in 2016?
✓ Correct — ✅ The investigation found the score was significantly more likely to incorrectly flag Black defendants — a pattern that could affect bail and sentencing decisions.
The investigation found the score was much more likely to label Black defendants as high risk incorrectly — a disparity affecting real decisions about people's freedom.
🧪 Why can't the COMPAS fairness dispute be solved just by making a better algorithm?
✓ Correct — ✅ The two fairness definitions that ProPublica and the company used are mathematically incompatible — you can't satisfy both at once. That makes it a values question, not just a technical one.
The two fairness definitions (equal false positive rates vs. equal predictive accuracy) are mathematically incompatible. Choosing between them is a values decision, not a technical fix.
🧪 Why isn't "80% accurate overall" enough information to judge an AI system?
✓ Correct — ✅ Average accuracy hides distributional problems — the same 80% overall could mean 90% for one group and 65% for another.
Overall accuracy averages across all groups. If errors are concentrated in one community, those people experience much worse accuracy even as the headline number looks fine.
🧪 What questions should you ask about any AI used in government decisions?
✓ Correct — ✅ Accountability questions: who made it, how was it trained, is it equally accurate across groups, and can people affected by it challenge its decisions?
The accountability questions: who built it, what was it trained on, is it equally accurate for all groups, and can affected people understand or challenge its decisions?
Explore what "fair" actually means — and why different people disagree.
Lab 3 — The Fairness Test
Fairness sounds simple, but it turns out there are different ways to define it — and sometimes they contradict each other. In this lab, you'll explore a scenario and figure out what "fair" would mean in different ways.
The AI will describe a simple scenario involving an algorithm making decisions about two groups.
You'll be asked: what would be fair? You give your answer.
The AI will show you a different definition of fairness that leads to a different answer — and you'll discuss which you'd choose and why.
There's no single right answer. The goal is to understand that "fair" isn't as simple as it sounds.
🔬 Lab AssistantLab 3
Here's the scenario: A school uses an AI system to decide who gets extra tutoring. Group A students score correctly 9 out of 10 times — the AI almost always identifies when they need help. Group B students score correctly only 7 out of 10 times. The AI's creators say "overall accuracy is 80%, which is fair." A parent from Group B disagrees. Who do you think is right — and what would "fair" mean here?
AI is helping doctors and scientists do things that were impossible before — but it's also showing us some surprising ways it can go wrong.
Amara wants to be a doctor. She reads about AI that can look at photos of skin and detect cancer earlier than human doctors. She's excited — that could save lives! But then she reads the fine print: the AI was trained mostly on photos from lighter-skinned patients. For darker-skinned patients, the accuracy was significantly lower.
Amara sits with that for a while. If she were the patient, which would she want — a tool that works great for most people, or one that works equally well for everyone? And if she were developing the tool, how would she make sure she got data from all kinds of patients, not just the easiest ones to collect?
What AI Has Genuinely Changed in Medicine
Some AI medical breakthroughs are real and significant. DeepMind's AlphaFold solved the protein folding problem — figuring out how proteins fold into their 3D shapes — that had stumped biologists for 50 years. This matters enormously for drug discovery and disease research. AI systems can detect certain types of cancer in medical images earlier and more accurately than human radiologists, in specific settings where they've been well-trained and validated.
These are genuine wins. But they come with important asterisks: the AI works well in the specific conditions it was trained for. Change those conditions, and performance can drop dramatically.
🔑 The Training Data Problem
An AI is only as good as the data it learned from. If a medical AI was trained mostly on images from one type of patient population, it may not work as well for people who weren't well-represented in that data. This isn't a small problem — if a cancer detection AI is less accurate for certain groups, those patients are at higher risk of missed diagnoses. Who gets included in the training data is a critical decision.
AI in Climate and Environmental Science
AI is also being used to tackle some of the biggest problems facing humanity. Climate scientists use AI models to analyze massive amounts of satellite, ocean, and atmospheric data to better understand how climate change is happening and where. AI is helping ecologists track endangered species, identify illegal deforestation, and monitor ocean health.
There's an irony here: AI systems themselves use a lot of energy and produce carbon emissions. Training a large AI model can emit as much carbon as several transatlantic flights. So the same technology being used to fight climate change is itself contributing to it — a tension researchers are actively working to address.
🌟 The Validation Question
Before any medical AI gets used on real patients, it needs to be tested carefully — not just on one dataset from one hospital, but across many hospitals, many types of patients, and many different conditions. This is called validation. A study showing high accuracy from one institution isn't enough. Good validation asks: does this work equally well for everyone it will be used on?
🧪 What did AlphaFold accomplish that had stumped scientists for 50 years?
✓ Correct — ✅ AlphaFold solved protein structure prediction — figuring out a protein's 3D shape from its sequence — a fundamental problem in biology that had resisted solution for fifty years.
AlphaFold solved the protein folding problem — predicting a protein's 3D shape from its chemical sequence — which has major implications for drug discovery and disease research.
🧪 Why was the skin cancer detection AI less accurate for darker-skinned patients?
✓ Correct — ✅ Training data that doesn't represent all patients means the AI learns patterns specific to the represented groups — and performs worse for those not well-represented.
An AI trained mainly on one population learns patterns specific to that population — and performs worse for groups that weren't well-represented in the training data.
🧪 What does "validating" a medical AI system mean?
✓ Correct — ✅ Validation means rigorous testing across diverse settings — not just one dataset from one hospital, but many hospitals, many patient types, actively looking for where it fails.
Validation means careful testing across many different hospitals, patient populations, and conditions — not just showing it works once in one study.
🧪 What is the irony in using AI for climate science?
✓ Correct — ✅ Training large AI models uses significant energy and produces carbon emissions — creating a real tension when AI is used to study and combat climate change.
There's a real tension: AI is helping scientists understand climate change, but training large AI models requires significant energy and produces carbon emissions itself.
🧪 The most important question about who is included in AI training data is:
✓ Correct — ✅ Representation matters: if the training data doesn't include the kinds of people the AI will serve, those people are at risk of getting worse performance.
The critical question is whether the data represents the diversity of people the AI will eventually be used on — gaps in representation translate into gaps in performance.
Figure out what would make a medical AI study trustworthy — for everyone.
Lab 4 — Design a Better Medical AI Study
Imagine you're in charge of testing a new AI that detects a disease from medical scans. Your job is to design a study that would make doctors trust the results for all patients — not just some of them.
The AI will describe a flawed study and ask you what's wrong with it.
Together, redesign it to fix the problems.
The AI will ask: what would still be hard to get right, even in a better study?
Think about: who was included in the study? How were they selected? How many different hospitals and types of patients were tested?
🔬 Lab AssistantLab 4
Here's the flawed study: Researchers tested an AI that detects a lung disease from chest X-rays. They tested it on 5,000 scans — all from one large hospital in one city. The patients were 85% from the same ethnic group. The AI scored 92% accuracy. The researchers declared it ready for hospitals nationwide. What problems do you see with this study?
AI is changing what kinds of jobs exist and how work gets done — in ways that could be good for some people and hard for others.
Nina's dad drives for a delivery app. Last week, his account got flagged and he couldn't work for three days. He still doesn't know exactly why. He contacted customer support, got a form email, and then another form email. There was no person to talk to. An algorithm made the decision, and there was no clear way to appeal it.
Meanwhile, Nina's aunt just started using an AI assistant at her office job. It drafts emails, summarizes long reports, and helps her prepare presentations. She finishes her work in six hours instead of eight and uses the extra time to do more interesting projects. For her, AI has made work better. For Nina's dad, AI made work scarier and less stable.
Same technology. Very different experiences.
How AI Is Changing Work
AI is changing work in two main ways: it's automating some tasks (doing them instead of a person) and augmenting others (helping people do them better and faster). Often it does both — automating the routine parts of a job while helping with the more complex parts. The impact depends a lot on what job you have, how much your employer uses AI, and whether the gains go to workers or just to company profits.
Research consistently shows that people who use AI tools at work get more done. But productivity gains don't automatically mean workers get paid more or work fewer hours. Whether that happens depends on things like labor unions, company culture, and government policy — not just the technology itself.
🔑 Algorithmic Management
In gig economy jobs (delivery, rideshare, warehouse work), AI doesn't just route orders — it manages people. Algorithms set pay, assign jobs, score performance, and sometimes fire workers automatically when scores fall too low. Unlike a human manager, the algorithm has no judgment, context, or ability to hear your side of the story. Many workers have described feeling managed by a system that treats them as interchangeable parts rather than people.
Jobs of the Future
People worry a lot about AI "taking jobs." The honest answer is that AI will change what jobs look like more than it will eliminate all work. Some specific tasks will be automated. New types of jobs that didn't exist before will emerge — like the people who train AI systems, audit them for fairness, or help companies figure out how to use them responsibly.
History shows that major technology changes do create new kinds of work — but the transition can be hard for people in the specific jobs most affected. The key question isn't just "what jobs will survive?" but "who benefits from the change, and who is left behind — and what do we do about that?"
🌟 Nina's Dad and Nina's Aunt
They're both experiencing AI in the workplace. For Nina's aunt, it's been positive. For her dad, it's been stressful and unfair-feeling. Both experiences are real. Understanding why they're different — and what could make things fairer for people like her dad — is exactly the kind of thinking that future AI policy needs.
🧪 What is the difference between AI "automating" and "augmenting" work?
✓ Correct — ✅ Automation replaces human work; augmentation enhances it. Many jobs involve both — AI handles routine parts while helping people with the complex parts.
Automating = AI does the task instead of you. Augmenting = AI helps you do the task better. Many jobs involve both happening at the same time.
🧪 What is "algorithmic management"?
✓ Correct — ✅ Algorithmic management means AI handles the decisions normally made by human managers — pay, job assignment, performance scoring, and even firing — often with limited human oversight.
Algorithmic management: AI systems make the decisions a human manager would — assigning work, setting pay, scoring performance, and sometimes deactivating workers — without much human judgment involved.
🧪 Why didn't Nina's dad know why his account was flagged?
✓ Correct — ✅ Algorithmic decisions often lack clear explanations and appeal mechanisms — unlike human managers, the system provides form emails rather than reasons and conversations.
The algorithm made the decision automatically. Without a human manager, there's often no explanation for why and no clear path to appeal — just form emails.
🧪 Why doesn't AI productivity growth automatically mean workers benefit?
✓ Correct — ✅ Technology creates the productivity gain; policy, labor agreements, and power determine who captures it. The technology alone doesn't ensure workers benefit.
Whether AI productivity gains go to workers or company profits depends on labor market conditions and policy — not automatically on how much more productive AI makes workers.
🧪 What new kinds of jobs might AI create?
✓ Correct — ✅ AI creates entirely new roles — training, auditing, governance, ethics — that didn't exist before. Not all AI jobs are programming jobs.
New AI-era jobs include training AI systems, auditing them for fairness, helping organizations deploy them responsibly — roles that require different skills than just programming.
What should change so workers like Nina's dad are treated more fairly?
Lab 5 — Fairer Rules for Gig Work
Nina's dad was flagged and couldn't work for three days, with no explanation and no appeal. Work with the AI to figure out what rules would make algorithmic management fairer — and why those rules don't exist yet.
Start by describing one rule you think should exist for platforms that algorithmically manage workers.
The AI will explain why the platform might push back against that rule.
Together, figure out what a fair compromise might look like.
Think about: what does a worker need to be treated fairly? What does a platform need to operate efficiently? Can both be true at the same time?
🔬 Lab AssistantLab 5
Nina's dad was suspended for three days with no explanation and no appeal process. If you could add one rule that all gig platforms must follow when their algorithm flags a worker, what would it be — and why? I'll tell you how the platform would likely respond.
AI can create images, music, and stories — raising new questions about creativity, ownership, and what makes something genuine.
Lena has been making digital art for two years. She spent hours learning to draw faces, studying light and shadow, experimenting with color. Then an AI image generator appeared that could produce beautiful artwork in seconds. At first she was angry. Then she tried it herself — and found herself using it to generate backgrounds and textures while she focused on the characters and emotion. She discovered she could make things she couldn't make before.
But a question kept nagging at her: the AI was trained on millions of artists' work — including, probably, artists who are now struggling to sell their own work because clients can get something similar from AI for free. Was she part of the problem? Was the tool itself fair?
What AI Can (and Can't) Create
Modern AI can generate images, music, poems, stories, and code that can be impressively good. Text-to-image tools like Midjourney and DALL-E produce visual art from text descriptions. Music generation tools create melodies and full tracks. AI can write in the style of specific authors or artists.
What AI can't do is create from personal experience. An AI doesn't know what it feels like to be scared, to fall in love, to lose someone, or to watch a sunrise. It generates combinations of patterns from what humans have created. That's genuinely impressive — but it's also different from human creativity in ways that matter to many people.
🔑 The Training Data Issue
AI art and music generators were trained on human-created work — often scraped from the internet without asking the artists or paying them. Those artists' labor is baked into every image the AI generates. Artists have filed lawsuits, arguing this is unfair. The legal questions aren't fully resolved yet. But the fairness question feels pretty clear to many people: if your work taught the AI, shouldn't you get something from that?
Deepfakes and Authentic Voices
AI can now clone voices from just a few seconds of audio, and create video of people saying or doing things they never said or did. These are called deepfakes. Most commonly, deepfakes are used to create fake content of real people — often targeting women and girls with non-consensual content, which is a serious harm. They're also used to spread disinformation — fake videos of politicians or celebrities.
Beyond the harms they cause, deepfakes create a trust problem: if you can't trust a photo or video to be real, how do you verify anything you see? Some researchers worry that as deepfakes spread, people will start using "it could be a deepfake" as an excuse to deny real, authentic evidence of things they don't want to believe.
🌟 Does the Process Matter?
Lena's question — does it matter how art was made? — is one people have asked about every new art technology. Photography was once called "not real art." Computer-generated graphics faced the same skepticism. The tools keep changing. What stays constant is that art, at its best, communicates something genuinely human. Whether AI can do that — or whether it's always just patterns that look like something human — is a question worth sitting with.
🧪 What is the main fairness concern about how AI art generators were trained?
✓ Correct — ✅ AI art generators learned from human-created work that was scraped from the internet — often without the artists' knowledge or compensation.
The key concern: AI art generators were trained on artists' work scraped from the internet without permission or payment to the artists whose work was used.
🧪 What is a deepfake?
✓ Correct — ✅ Deepfakes are AI-generated synthetic media — video, audio, images — that portray real people saying or doing things they never actually said or did.
Deepfakes are AI-generated synthetic content — typically video or audio — that makes real people appear to say or do things they never did.
🧪 What is the "liar's dividend" problem with deepfakes?
✓ Correct — ✅ The liar's dividend: knowing deepfakes exist gives bad actors a way to dismiss real, authentic evidence — "that video is probably fake."
The liar's dividend: the existence of deepfakes gives people a built-in excuse to reject authentic evidence they don't want to believe — "that video could be AI-generated."
🧪 What can AI-generated art NOT do that human artists can?
✓ Correct — ✅ AI generates combinations of patterns from human-created work — it doesn't have personal experiences, emotions, or a life story to draw from.
AI creates by combining patterns from what humans made — it has no personal experiences, fears, joys, or losses to draw from. That's a fundamental difference from human creativity.
🧪 How did Lena end up using AI in her artwork?
✓ Correct — ✅ Lena found a way to use AI as a tool that extended her creative capability — using it for elements she could outsource while focusing her personal skill where it mattered most.
Lena found AI useful as a tool: generating backgrounds and textures while she focused her own creative energy on characters and emotional expression.
Work through what makes something genuinely creative — and where you draw the line.
Lab 6 — Is This Real Art?
Different people draw the line in different places about what counts as "real" creative work when AI is involved. In this lab, you'll think through your own position.
The AI will give you four short descriptions of how different pieces of content were created — some more AI-involved than others.
For each one, decide: does this count as the person's creative work?
The AI will ask you to explain what makes each decision different — to find the principle you're actually using.
You're looking for consistency: a principle that explains all your decisions, not just each one separately.
🔬 Lab AssistantLab 6
Here are four ways someone could create a piece of artwork. For each one, tell me: is this the person's creative work?
1. A painter uses brushes and paints to create an original piece from their imagination.
2. A photographer takes a photo and then edits the colors and composition on their computer.
3. A designer uses an AI tool to generate a starting image, then redesigns and repaints major parts of it digitally.
4. Someone types a text prompt into an AI generator and picks the best result from twenty options.
Start with the one you feel most certain about.