1. MIT Sloan research found that "accuracy motivation" outperforms specific detection techniques in resisting misinformation. The key reason is that detection techniques are most often applied:
Correct. Selective application of skepticism means the techniques fail where they're most needed. Accuracy motivation corrects this by prompting consistent questioning regardless of whether content is agreeable.
The problem isn't speed or expertise — it's that people apply their knowledge inconsistently, sparing content they want to believe from scrutiny.
2. The Yale study found that asking "Is this accurate?" reduced false headline sharing by 37%. This primarily suggests that misinformation spreads because:
Correct. The interface design, not individual intelligence, is the primary structural factor. A single question inserts the evaluation step that the design omits.
If a one-word question reduces false shares by 37%, what does that imply about why those shares were happening in the first place?
3. The lesson says the five-question checklist will eventually be passable by sophisticated AI when prompted carefully. What is the lesson's recommended response to this limitation?
Exactly. The lesson explicitly separates the checklist (specific indicators with a shelf life) from the underlying question (which is more durable). The goal is to hold onto the question, not the current list of answers to it.
The lesson ends by saying "not the checklist, but the question the checklist is pointing at." The underlying question — did thinking actually happen? — is what persists across changing AI capabilities.
4. AI image generators produce errors in hands because:
Correct. High variation in training data means the AI averages poorly — it never locks onto a reliable template for hands.
Think about what AI image generators learn: statistical patterns. What happens when the pattern varies enormously across examples?
5. What does the MIT 2018 study's finding — that false news spreads faster because it is more novel and emotionally triggering — tell you about your own verification habits?
Correct. If false content is engineered to trigger sharing before thinking, then the urgency you feel to share is itself a warning sign. Developing the habit of pausing when you feel that urgency — and running at least a quick verification — is one of the most valuable outcomes of this module.
Emotional reactions don't mean content is false — real tragedies and genuine injustices also produce strong feelings. But the urge to share immediately is precisely the reaction that misinformation is designed to trigger. Treating urgency as a reason to slow down is a practical habit that reduces vulnerability significantly.
6. The 2023 AI-generated Pentagon explosion image moved financial markets before being debunked. Which of the following best explains why individual verification skills couldn't have prevented this outcome at the system level?
Correct. Trading algorithms and social media amplification respond in milliseconds. A five-minute manual verification process — however accurate — cannot prevent consequences that occur in the first ninety seconds of a fake image's spread.
The problem isn't detection difficulty — the image had multiple detectable flaws. The problem is speed: automated systems act on information faster than human verification can run. That's a structural challenge that individual skill alone can't solve.
7. The AI-generated image of Pope Francis in a puffer jacket went viral in February 2023. What was the creator's name and original platform?
Correct. Pablo Xavier, a Chicago construction worker, created and posted the image in a Facebook hobby group — from which it escaped into the wider internet.
Pablo Xavier, a Chicago construction worker, created the image and posted it in a Facebook hobby group. It spread from there without his labels.
8. A real construction site photo has shadows that appear slightly inconsistent because of a reflective glass building nearby bouncing light. An analyst applying the lighting checklist should:
Correct. Good forensic analysis eliminates alternative explanations. A glass building creating secondary reflections is a real physical cause for shadow anomalies — one that removes this from the AI artifact column.
Forensic analysis requires eliminating alternative explanations. If reflected light from a glass building explains the shadow anomaly, that anomaly is not an artifact — it has a real physical cause. Artifacts are anomalies with no other plausible explanation.
9. What was the key insight from the Amazon fake review detection story that justified building a checklist based on "whether thinking happened"?
Right. The arms race insight: specific patterns become patch targets. The underlying question — did someone actually think this through? — is more durable because it points at a structural reality, not a surface signal.
The Amazon engineer's insight was specifically about why pattern-based detection fails: it's gameable. The more durable question is the one that asks about the underlying reality patterns are trying to point at.
10. Stanford and MIT researchers found that readers rated a fluently-written incorrect fact as more credible than an awkwardly-written version of the same incorrect fact — even when told both had the same accuracy. What does this tell us about fluency bias?
Exactly. Even when readers consciously knew both versions were equally accurate, the fluent version still won out. The bias runs deeper than conscious reasoning — which is what makes it dangerous in real reading situations.
The critical finding is that the warning didn't eliminate the bias. Readers knew the versions were equally accurate but still rated the fluent one higher. This shows the effect runs below conscious evaluation.
11. A GAN trains two AI systems against each other. The end result of this competition is:
Correct.
The competition drives both systems — but the generator's output is what reaches the public and keeps improving.
12. Three deepfake detection areas where current models most often fail are:
Correct.
These specific three areas are where the lesson identified current model weaknesses — review Lesson 1's detection section.
13. Metadata can be edited after a photo is taken. What does this mean for how you should use metadata in verification?
Correct. The editability of metadata doesn't make it useless — it makes it one piece of evidence rather than definitive proof. Consistent metadata increases confidence; contradictory metadata is a significant red flag. Neither is conclusive without corroboration from other methods.
Metadata is valuable but not infallible. Its editability means it must be used alongside other verification methods. Multiple independent lines of evidence all pointing to the same conclusion — triangulation — is what builds reliable confidence.
14. What is a "catchlight" and why does it matter for AI image detection?
Correct. Catchlights are the light source reflections in the eyes. In real photos, both eyes catch the same source. AI images sometimes show different reflections in each eye, or reflections of light sources that don't exist anywhere in the scene.
A catchlight is the reflection of the light source visible in a person's eyes. Because both eyes see the same scene, a real photo shows matching catchlights. AI images sometimes produce different or impossible reflections in each eye.
15. You receive a video of a local official making a statement you find deeply surprising and politically important. You want to share it immediately. Based on everything in this module, what is the most responsible single first step?
Correct. Systematic sourcing — tracing the chain of custody — is identified in the lesson as the most durable detection skill. It works even as specific technical artifacts improve in future models, because authentic footage has a traceable origin that synthetic content typically doesn't.
Technical checks (hairline, blinking) are useful but expire as models improve. Sourcing — where did this first appear — is the most durable first step identified in the lesson.
16. According to research cited in this module, what percentage of deepfake videos online in 2023 were non-consensual intimate imagery?
Correct. The Sensity AI 2023 estimate was over 96% — a figure that reframes the common narrative about deepfakes being primarily a political problem.
The figure was significantly higher than most people assume — over 96% according to Sensity AI research cited in Lesson 1.
17. EXIF data in a photograph typically contains all of the following EXCEPT:
Correct. EXIF data records technical and location information about how and where an image was captured — not about who is depicted in it. Facial recognition is a separate technology entirely.
EXIF data covers technical metadata: when, where, and with what device an image was captured. It does not identify people in the photograph — that requires facial recognition, which is a completely separate system.
18. InVID / WeVerify is primarily designed for which verification task?
Correct. InVID's primary function is frame extraction from video, converting an unreverable-searchable clip into a series of individual images that can each be run through reverse image search engines.
InVID's core capability is breaking videos into key frames for individual reverse image searches — solving the fundamental problem that video clips can't be directly reverse-searched the way still images can.
19. The March 2022 Zelensky deepfake was debunked quickly. Why does quick debunking not fully neutralize the harm of such a video?
Correct. The target audience for the deepfake was narrow — soldiers and civilians in a moment of crisis — and even brief confusion in that audience could have real effects, regardless of what general audiences eventually concluded.
Think about who the intended audience was and what a short window of uncertainty could accomplish in that specific context.
20. Temporal inconsistency in deepfake video is best described as:
Correct. Temporal inconsistencies are inter-frame artifacts — they appear between frames rather than within them, and slowing to 0.25x speed makes them visible.
Temporal means across time — these are errors between frames, not within a single frame. Review the three detection layers from Lesson 4.