People judge frequency and probability by how easily examples come to mind. Memorable, vivid, or recent items seem more common than they are. This heuristic is often accurate but produces systematic biases when availability is influenced by factors other than actual frequency such as salience or media coverage.
From your overview of cognitive biases, you know that heuristics are mental shortcuts — efficient rules of thumb that generally serve us well but can produce predictable errors under specific conditions. The availability heuristic, described by Kahneman and Tversky in 1973, captures a specific shortcut: when estimating how common or probable something is, people don't count examples from memory — they judge by *how easily examples come to mind*. If you can quickly generate several instances of something, you conclude it must be common. If examples come slowly or not at all, you conclude it must be rare.
The heuristic works reasonably well because, in most natural environments, frequent events *are* easier to recall than rare ones — you've encountered them more often, so they've been encoded more times and have stronger memory traces. Ask someone to estimate how often they've driven versus flown, and availability will yield an accurate answer: driving examples flood in, flying examples are sparse. But availability is influenced by factors entirely unrelated to actual frequency — and that's where systematic bias creeps in. Vividness makes events feel more available: a dramatic plane crash is more memorable than an equally fatal series of car accidents, so plane travel feels more dangerous than driving despite the statistics reversing this impression. Recency inflates estimates: a friend's recent illness makes disease feel more prevalent than before you heard about it. Media coverage is perhaps the most powerful distorter: events that receive saturating news attention — shark attacks, terrorist bombings, lottery wins — become highly available and therefore feel far more probable than base rates warrant.
A classic demonstration is the "K word" study: participants asked whether more words in English begin with the letter K, or have K as the third letter, overwhelmingly choose "begin with K" — because words starting with K flood in easily (king, kind, key) while third-letter K words are hard to generate (ask, awkward, ankle). In reality, third-letter K words are about twice as common. Nothing about the actual frequency is changing; only the ease of retrieval differs. This illustrates that availability isn't just a rough proxy for frequency — it can be systematically inverted when retrieval dynamics favor one category over another.
The bias has significant real-world consequences. Risk perception research consistently shows that people overestimate deaths from dramatic causes (tornadoes, plane crashes, murder) and underestimate deaths from undramatic but common causes (diabetes, stroke, falls in the elderly). This misperception distorts policy priorities, personal decisions, and jury judgments. Importantly, availability is not purely a memory phenomenon — Norbert Schwarz showed that the *experience of difficulty in retrieval* also matters. When asked to recall twelve instances of assertive behavior (which is hard), people rate themselves as less assertive than when asked to recall six — the ease of generating six examples communicates "I'm an assertive person," while the struggle to generate twelve communicates the opposite, even though more examples were retrieved.