Abstract:

Goldstein & Gigerenzer (2002, p. 76) define the recognition heuristic thus: "If one of two objects is recognized and the other is not, then infer that the recognized object has the higher value with respect to the criterion." The idea behind the recognition heuristic is that, given how human cognition works and how society is structured, we're more likely to encounter "big" things (on whatever criterion dimension) than "little" things. The recognition heuristic is thus best understood as an instance of scaffolded cognition (Sterelny 2010). Engineering our cognitive environments to scaffold our cognitive 
capacities is a double-edged sword; with increased capabilities come increased vulnerabilities. There are obvious benefits to distributing one's cognitive load onto the world: storing contact information in 
smartphones, using cross platform calendars for scheduling, and communicating in searchable e-mail all serve to decrease internal processing demands while increasing access to information. Moreover, 
technological artifacts are not the only worldly items that play this role: we regularly rely on the reports of other people to decrease our own cognitive processing load while increasing our access to 
information. Cognitively speaking, however, there are no free lunches. Decreased cognitive demands and increased access come at the price of epistemic vulnerability. Gigerenzer & Goldstein (1996, p. 651) refer to the city-population task as their "drosophila," so if the recognition heuristic works anywhere, it should work here. I replicated some of the basic research on the recognition heuristic in an international setting by generating a dataset from which both ecological and surrogate correlations could be calculated. Results suggest that, after correcting for Zipf's Law, the recognition heuristic is internationally robust; however, its reliability is diminished outside of Europe, pointing to a kind of epistemic injustice. This suggests that, even when the speaker is a cooperative and the hearer is epistemically just in Miranda Fricker's (2007) sense, things can still go wrong. One primary way in which things can go wrong is that the speaker's message can get garbled or completely silenced by the media on its way from the speaker to the audience. Even people who conscientiously read the New York Times every day for a decade will be systematically misinformed about the world in which they live. The Frickerian model of communication as a direct speaker-to-hearer relationship is therefore inadequate. We need instead a three-constituent model that includes (potential) speakers, the media, and (potential) hearers, and which assigns distinct virtues to all of them.