Side-by-side in one pass
Both availability and representativeness answer hard probability questions with a simpler attribute. Availability substitutes ease of retrieval or imagination. Representativeness substitutes similarity to a mental model. Tversky and Kahneman mapped both in the early 1970s work that later sat inside the 1974 Science review Judgment under Uncertainty: Heuristics and Biases. That same review also covers anchoring and adjustment, which is useful as a third column because feeds and headlines often combine all three.
Neither shortcut is "stupid." They are fast defaults. They fail when the attribute you substituted (vividness, resemblance, or a starting number) is a poor stand-in for the frequency or probability you actually needed.
Comparison table
Three objects from the same research line, scored on the same dimensions so the differences stay mechanical rather than rhetorical.
Dimension | Availability | Representativeness | Anchoring and adjustment
Core question you replace | "How often / how likely?" becomes "How easily can I bring an example to mind?" | "How likely?" becomes "How much does this match the stereotype or process?" | "What value?" becomes "Start near this number, then nudge"
Classic cue | Recent headlines, dramatic anecdotes, easy-to-picture scenarios | A personality sketch that "fits" a job, diagnosis, or cause | A posted statistic, list price, poll seed, or first comment figure
Typical error | Overweight rare but memorable events; ignore quiet base rates | Ignore base rates and sample size when the story fits | Final estimates stay too close to an arbitrary start point
Media tell | One viral clip outranks a dull statistic in the same article | A profile that matches a villain or hero template feels like proof | The first number in a thread becomes the debate's center of gravity
Primary sources on this page | 1974 Science review (availability section) | 1972 Cognitive Psychology paper + 1974 review | 1974 Science review (anchoring section)
Availability: ease feels like evidence
Tversky and Kahneman's availability work showed that people judge frequencies by retrieval fluency. Events that are dramatic, recent, or heavily covered feel common even when they are not. A plane crash saturates a news cycle; routine car fatalities do not. The mind answers with the footage it can play back.
The 1974 Science paper walks through constructions such as judging letter frequency from how easily words come to mind, and judging risk classes from how quickly instances can be retrieved. The mechanism is not "people like drama." It is that retrieval fluency is treated as a proxy for frequency.
Online, availability is why one shared anecdote can beat a linked study in the comments. The anecdote is easier to simulate. Media literacy here is mechanical: before you update a belief from a story, ask what class of events is hard to retrieve because it is boring. Outrage that spreads because it is easy to recall sits next to half-life of online outrage.
A second tell from the same research line is unequal coverage inside a class: if one subtype of accident, crime, or medical scare is filmed constantly while another subtype is not, availability will inflate the filmed subtype even when the quieter subtype is larger in official counts. The fix is not to ignore vivid cases. It is to ask what the retrieval set is missing before you treat fluency as a frequency estimate.
Representativeness: likeness feels like probability
Representativeness is the similarity shortcut. If a description sounds like a librarian stereotype, people raise the odds the person is a librarian even when the base rate of librarians is tiny relative to other occupations. Kahneman and Tversky's engineer-lawyer style problems made the same point with base rates that participants largely ignored when a sketch felt diagnostic.
Their 1972 Cognitive Psychology paper on subjective probability treats representativeness as judging an event by how much it resembles its parent population or generating process. The 1974 review then places that shortcut beside availability and anchoring for a general audience. The failure mode is not "stereotypes exist." It is treating resemblance as if it already included the base rate and the reliability of the description. A profile that fits can still be rare.
Conjunction-style errors in the same program show the same likeness pressure: a detailed story that "fits" can feel more probable than a simpler inclusive event, even though the detailed story is logically narrower. That is why a neat narrative in a news package can feel stronger than a blunt base-rate sentence that is actually wider and safer.
Base-rate neglect in argument form shows up beside deductive vs inductive reasoning.
Anchoring as the third shortcut
Anchoring and adjustment, as summarized in the 1974 Science review, starts people at an initial value and leaves final estimates too close to that start even when the anchor is arbitrary. In media arguments the anchor is often the first number someone posts: a casualty figure, a salary, a poll margin, a "scientists say X%" claim without the denominator.
Availability and representativeness can feed the anchor. A vivid clip (availability) that also "fits the type" (representativeness) arrives attached to a round number that then becomes the adjustment floor. Separating the three questions (What is easy to recall? What matches the stereotype? What number arrived first?) keeps the diagnosis from collapsing into a single vague "bias" label.
How to tell which one you are using
If your confidence comes from how fast an example popped up, you are in availability territory. If it comes from how neatly a case matches a type, you are in representativeness territory. If it comes from refusing to leave a first number, you are in anchoring territory. Many hot takes use more than one at once.
A practical check for any of them: write the base rate, the sample size, or the source of the starting number in one sentence before you share the conclusion. If you cannot, you are still on the substitute attribute. For related bias about sticking with a first belief, see belief perseverance.