Module Check: Epistemics — Knowing in an AI World
This lesson is your consolidation point for the entire module. You have covered the classical foundations of knowledge — justified true belief, evidence, calibration — and traced their implications into an AI-saturated world where hallucination, homogenization, automation bias, and the myth of AI neutrality are live epistemic threats. Before the flashcards and quizzes, take a moment to ask yourself: what has actually changed in how you think about knowing? The goal of this module is not a list of terms but a genuine upgrade to your epistemic operating system.
Flashcards — click each card to reveal the answer
Module Quizzes
Kenji believes that a specific drug is effective because his friend told him a doctor said so, and the drug really is effective according to clinical trial data. Which element of the JTB account is most clearly strained in Kenji's case?
A food blogger with five million followers confidently claims that a specific dietary intervention cures a chronic condition, citing their personal transformation. A team of clinical researchers publishes a systematic review of 40 randomized controlled trials finding no statistically significant effect. Which source should receive substantially more epistemic weight, and what is the primary reason?
A superforecaster maintains a prediction journal and reviews their forecasts quarterly. They find that over 200 predictions rated '90% confident,' they were correct 72% of the time. What should they conclude and do?
An AI assistant generates a detailed paragraph about a niche historical event, including specific dates, participant names, and a citation. The paragraph is fluently written with a confident tone. What is the epistemically appropriate response before using this content in an academic paper?
An education company deploys a single AI tutoring system for 10 million students across all subjects. Over several years, teachers notice that students in different schools, cities, and countries produce remarkably similar frameworks and vocabulary when discussing open-ended questions in history and ethics. Which epistemic risks are most directly implicated?
A student reads a highly confident AI summary of a contested scientific debate and concludes: 'I can see both sides now and understand the controversy.' What critical epistemic limitation of this approach has the student missed?
Synthesis: Your Epistemic Commitments
Epistemology is practiced, not just known. The concepts in this module are tools — and like all tools, they must be used repeatedly and deliberately to become reliable. The gap between knowing that calibration matters and being well-calibrated is bridged only by practice: keeping predictions, seeking disconfirmation, auditing your beliefs, and correcting your track record. The AI dimension of this module is not about fear or dismissal of a powerful technology. It is about using that technology with clear-eyed understanding of what it is good at and where it breaks down. AI can dramatically extend your research capacity, surface ideas you would not have encountered, and help you think through complex problems. It can also hallucinate with perfect confidence, homogenize your thinking if you let it, and create dependence that atrophies your independent judgment if you do not actively resist. The synthesis activity below is your opportunity to put everything together.
Synthesis: The Epistemics Design Brief
- You are advising a team designing an AI-assisted research tool for high school students. Your job is to write a one-to-two page Epistemic Design Brief that specifies how the tool should handle the epistemic risks you have studied, and what epistemic habits it should build in its users.
- Your brief must address each of the following:
- 1. Hallucination mitigation: What specific design choices would reduce the harm from hallucination? Consider what information the tool should always flag, what verification prompts it should build in, and what claims should always require external sourcing.
- 2. Calibration communication: How should the tool represent its own uncertainty? Should it ever give a confident answer without hedging? What language should it use to communicate the difference between high-confidence and low-confidence outputs?
- 3. Epistemic diversity preservation: What design choices would prevent the tool from homogenizing student thought? Should it ever withhold its own synthesis? Should it present competing frameworks? How does it encourage students to form their own views?
- 4. Anti-dependence features: What would you build in to ensure students develop independent reasoning capacities rather than substitute AI for their own thinking? What would you NOT allow the tool to do, even if technically possible?
- 5. Authority and trust transparency: How would the tool communicate its own limitations, biases, and the boundaries of its domain reliability? What should users always be told when they ask it something?
- For each design choice, explicitly name the epistemic concept from the module that motivates it.
- Present your brief to a partner or the class and discuss: where do your designs differ, and what do those differences reveal about your epistemic priorities?