Abstract
How many faces can a person recognize? We consider three classes of relations where we have an opportunity to learn a person's face: a) celebrities (e.g., entertainers, athletes, politicians, rogues) with which we have no personal contact, b) individuals with whom we have personal contact such as friends, frequent service people (e.g., physicians, instructors), classmates, coworkers, and family, many of whom we can name and cite biographical details, and c) people we know from a specific context, such as a barista or a passenger on our usual bus route, for whom we know only from that context, rarely by name or biographical detail. To assess celebrity recognition, participants were presented with a pair of faces and asked to judge which one is famous. One was a headshot of a celebrity sampled at various degrees of fame (US Weekly Estimate) and the other a non-celebrity doppelganger of the celebrity (Fig. 1). The integral of the accuracy function over citation frequency can provide an estimate of the first class (recognizable celebrities) (Fig. 2a). The second class (personally familiar), will be estimated from the number of Facebook friends that the subject confirms are recognizable plus the cumulative number of names or identifying descriptions generated over a two-week period, excluding those already accounted for from Facebook (Fig. 2b). The third (contextually familiar) will be estimated from logs that participants will be asked to keep for two weeks in which they mark the number of familiar faces that they encounter each day without knowing the names of those individuals. The asymptote of the number of unique individuals encountered will be the estimate predicted from the two-week sampling (Fig. 2c). The sum of the three estimates will provide, for the first time, a lower-bounds estimate of the number of identifiable faces.
Meeting abstract presented at VSS 2018