When Tversky and Kahneman introduced the availability heuristic in 1973, they described a simple mental shortcut: we estimate the probability of an event based on how easily examples spring to mind. Plane crashes feel more dangerous than car crashes partly because they dominate news coverage — they are cognitively available, even though statistically you are roughly 95 times more likely to die in a car. The heuristic worked adequately in small social groups, where memorable events genuinely tended to be frequent ones. It was never designed for a world of algorithmically curated information.

A 2024 study published in Nature Human Behaviour tracked 4,000 participants across six countries and found that heavy social media users systematically overestimated the prevalence of violent crime, rare diseases, and economic collapse — not because they were misinformed, but because their feeds had optimised for engagement, and fear and outrage generate the most clicks. The algorithm had weaponised availability.

What makes this especially concerning is a feedback loop researchers now call algorithmic amplification of availability bias. A user who dwells slightly longer on a crime story is shown more crime stories. Their subjective sense of how common crime is rises. They engage more with alarming content. The platform's model interprets this as interest and doubles down. Within weeks, a mild initial skew can become a profoundly distorted worldview — built entirely from genuine news, just deeply non-representative news.

Fun fact: After a shark attack receives major news coverage, Google searches for "shark attack" spike — but so do searches for "is swimming safe?" Researchers have used this as a real-time proxy for availability bias in the wild, tracking how media cycles inflate perceived risk for up to three weeks after a prominent story.

Interestingly, a 2025 preprint from MIT Media Lab found that simple base-rate prompts — brief on-screen reminders showing the actual statistical frequency of events before a user reads about them — reduced availability-driven misestimation by 31% without reducing engagement. The intervention required no behaviour change from the user; the platform simply had to contextualise the story.

The availability heuristic also operates in quieter, more personal ways. Hiring managers who recently interviewed a strong candidate raise their standards — recent exemplars become the mental anchor. Doctors who have just seen a rare disease are more likely to diagnose it in the next patient. Investors who lived through a crash overweight crash probability for years after. In each case, the mind is not broken — it is doing exactly what it was designed to do, reaching for the most cognitively vivid reference point. The design just predates the internet.