iSee platform is coming to life this year. The team has built a Web application of the Cockpit for user case design and utilisation. We start using our cockpit to model the 12 real world use case identified with our industry partners – Jiva, BT, Bosch, TotalEnergies, AutoMix and AAI-MX lab.
The Cockpit is being wired up with a conversation layer (in progress) which will allow utilise iSee to get recommendation from AI model coupled with explanations and questions for evaluating the quality of the explanation.
One of the benefits of iSee is to discover reusable explanation strategies across multiple cases and goals. Thus we first need to have a robust framework to represent what we mean by “explanation strategy” in a format that can be processed through similarity measures and translated into a conversation script/dialog meaningful for the end-user. In iSee we decided to use Behaviour Trees and thus we extended a tool for editing and building behaviour trees customised to represent explanation methods and evaluation questionnaire.