Drawing on philosophy and relational practice, these views inform how we build AI – and what we ask it to contribute.
We bring expectations, commitments and a position in the situation. These affect what we notice, what we take for granted and what we do next. Understanding can deepen when we recognise how our own participation is shaping what happens.
The same words can carry different expectations. “Support”, “agreement” or “responsibility” may mean something different to each person involved. Understanding develops as those meanings meet, become clearer and sometimes change. One person’s account is a starting point, never the whole situation.
People can recognise a doubt before they can explain it. A situation can change while they are making sense of it. What helps at one moment may settle things too soon at another. AI needs to respond to that movement without treating every emerging thought as a finished conclusion.
A question directs attention. An interpretation offers a way of seeing. An answer can provide useful ground – or make further thought feel unnecessary. AI participates through the responses it gives. That makes what it says, how and when a matter of judgment – and that judgment has to be tested.
What develops with Kwalia should help people beyond the conversation: in their work, their decisions and their encounters with others. Sometimes the most useful contribution is helping someone take the next step outside the exchange.
We’d like to hear from researchers and builders working on model behaviour, evaluation, and what AI must become next.