Fiche du document
26 octobre 2018
- handle: 10670/1.udoxom
- hal: hal-02333913
Citer ce document
Grazia Cecere et al., « Computer algorithms prefer headless women », HAL-SHS : études de genres, ID : 10670/1.udoxom
Métriques
Partage / Export
Résumé
Advertising algorithms power the online digital economy. However, it is unclear whether they may end up distorting the kind of information people are exposed to. To explore this we ran a randomized online ad campaign on Snapchat on behalf of a French computer science school that explored how the ad algorithm allocated pictorial content representing gender. Our results show that pictures depicting a complete male torso was shown more to teens, while the female picture that was displayed most by the algorithm depicted the woman as not having a head. We present suggestive evidence that these algorithms are driven by preferences in large population centers in Paris as it appears the algorithm determines which images are the most "engaging" on the first day for the places with the largest numbers of users and replicates this pattern going forward elsewhere.