L'étude de pertinence d'un nouveau logiciel Microsoft a été conduite selon la méthodologie nouvelle suivante, fondée sur l'"embauche" de centaines de "Mechanical Turks" :
Participants were paid Mechanical Turk workers in the United States.
Each pair of Turkers chatted for 10 minutes using: 1) CAHOOTS, our CGH system, 2) plain chat (no image suggestions), or 3) a CGH system with computer-generated images, all using the same interface In the CGH system, whenever one user sends out a message, our system automatically inserted the single top-ranking funny image suggestion into the chat, with "computer:" inserted above the message, as is common in systems such as WhatsApp. Assignment to system was based on random assignment.
A total of 738 participants (408 male) used one of systems, with at least 100 participants using each variant. Pairs of participants were instructed how to use the system and asked to converse for at least 10 minutes. After the chat, participants were asked to fill out a survey to evaluate the conversation and the system.
Encore un domaine (celui des testeurs sociologiques) que Internet va complètement bouleverser : finis les panels à la papa ...
