Lucia Dossin

Cassandra

Cassandra is a voice-operated chatbot, aimed at making psychological profiles during the conversations. These profiles are shown as pie charts where every color corresponds to a personality attribute. The graphics are accompanied by a short analysis and a list of companies possibly interested in the profile. In this sense, the bot provides data that can act as a mediator in services like job-seeking and dating. The profile then becomes a digital representation of the user and can be used as a kind of digital key to personalize every gadget in the near future of the Internet of Things. Make your life easier, let Cassandra make a digital copy of you.

Presented at Hofpoort as part of the exhibition Tempted by Tomorrow.

Images

Cassandra Card

Interface

Welcome Screen
Screen 2/3 - Chat
Screen 3/3 - Profile Results

Exhibition

The process

The pie chart graphic resulting from the conversation is composed of 6 colors - each one representing one personality attribute. The one closer to the center is based on the frequency in which the user says certain words and represents the degree of self-confidence. The other 5 colors are based on answers to yes-no questions, triggered during the conversation every time certain keywords are detected.

These keywords were defined through free association by a group of friendly volunteers. They were presented one word (which was taken from a psychological profile questionnaire and related to a certain personality attribute) and made the free association. For example: they were presented the word 'extremists', which can be found in the questionnaire relating to Social and Political Attitudes. Several connections were made to this word and they are all used to trigger questions that will evaluate Social and Political Attitudes during the conversation with the user.

This is the same mechanism I used in my previous experiment The Free Association Database.

Free Association - input screen and result sample
Words associations visualization - the interactive version can be found here.
Relationship between keywords and questions asked during chat - code sample