Technology. Human experience.
The conversations in between.
Help shape how we understand conversational AI. Bring a question, challenge a finding, contribute your expertise or support the next investigation.
Find your way inThree ongoing research threads, and the ways you can help move them forward.
HCIS studies how earlier context is understood, reused and reflected in later behaviour during sustained human–AI conversations. The framework and its protocols have been substantially rewritten; independent validation remains ahead. Reviewers and collaborators can help test the boundaries, apply the protocols to other conversations and examine whether reviewers reach consistent judgements.
What changes when a companionship judge can see the conversation around a reply? Anna’s INTIMA research note explores context-sensitive judgements in one long-running companion system. The next question is how these observations compare with other people’s results.
Read the research note ↗An ongoing investigation of J-space in AI companions. We will share the research direction in an upcoming blog post; collaborators working on related questions are welcome.
Research note coming soonYou don’t need to be a researcher.
You do need curiosity and care.
A replication, a thoughtful challenge, a practitioner observation or a question from your field. Help us see what we’ve missed.
For researchers, practitioners & curious mindsStart a conversationHave a specific project or complementary expertise? Tell us who you are, what you’d bring and what you’d like to explore together.
A short introduction is a good first stepIntroduce yourselfHelp make another evaluation, comparison or follow-up possible. Explore the research your contribution would support.
Small contributions are welcomeSee where tokens are goingTokens are the pieces of text a model processes. Research uses them to run conversations, compare responses and examine results.
Support helps create room for careful comparisons, independent scrutiny and follow-up work. It does not determine the findings.
Get in touch about gifting AI credits or sponsoring our research tools
Explore behaviour across defined scenarios and repeated interactions.
Compare judgements and investigate differences that deserve a closer look.
Test alternative explanations and make room for results that challenge expectations.
We agree acknowledgement and credit with contributors according to the work contributed. Please contact us before starting substantial work intended as a collaboration.
We have limited capacity for individual guidance. A brief introduction helps us assess fit. Please share public work or a summary of your idea; keep private conversations and third-party details out of your message.
[email protected] ↗