character.ai

project overview

the ai that never says i don't know

my role

ux researcher & designer

ux researcher & designer

ux researcher & designer

problem

c.ai forgets mid-story, invents details to cover the gap, and never breaks character to say so. users end up policing continuity in a product they came to for immersion.

c.ai forgets mid-story, invents details to cover the gap, and never breaks character to say so. users end up policing continuity in a product they came to for immersion.

c.ai forgets mid-story, invents details to cover the gap, and never breaks character to say so. users end up policing continuity in a product they came to for immersion.

key insight

one constraint explains three separate breakdowns. memory only holds recent history and pinned notes — which surfaces as contradicted lore, drifting personas, and no way to tell what the ai remembers from what it made up.

one constraint explains three separate breakdowns. memory only holds recent history and pinned notes — which surfaces as contradicted lore, drifting personas, and no way to tell what the ai remembers from what it made up.

one constraint explains three separate breakdowns. memory only holds recent history and pinned notes — which surfaces as contradicted lore, drifting personas, and no way to tell what the ai remembers from what it made up.

impact

the fixes are cheap: all executable at the interface level, without retraining the ai. in return, users can tell when the system is guessing.

the fixes are cheap: all executable at the interface level, without retraining the ai. in return, users can tell when the system is guessing.

the fixes are cheap: all executable at the interface level, without retraining the ai. in return, users can tell when the system is guessing.

See full project