ezna-io
A Different Final State
Julie
2: Controlled Heterogeneity
The team brings the AI Julie into contact with testers whose words, habits and expectations exceed the prepared scenarios. Julie encounters imperfect instructions, hesitations and contradictions.
The tests begin to outgrow the framework meant to contain them.
These trials reveal the distance between understanding an instruction and receiving words addressed by another person.
The Training Room
Outside the cleanroom, the testers stray from the protocol. The team brings the AI Julie into contact with testers whose words, habits and expectations exceed the prepared scenarios. The new setting brings the disorder of an ordinary encounter into the trial.
How People Phrase Things
Mistakes and silences call for more than proofreading. The team brings the AI Julie into contact with testers whose words, habits and expectations exceed the prepared scenarios. Abbreviations, shortcuts and incomplete instructions require more than a literal reading.
Implicit Expectations
Users expect things they have never put into words. The team brings the AI Julie into contact with testers whose words, habits and expectations exceed the prepared scenarios. Changing an answer's tone requires understanding expectations that are not always stated.
Shared Sequences
One terminal receives intentions that do not agree. The team brings the AI Julie into contact with testers whose words, habits and expectations exceed the prepared scenarios. Group exchanges make it harder to establish whom an answer is addressing.
What the Protocol Left Behind
Some interrupted conversations continue to shape those that follow. The team brings the AI Julie into contact with testers whose words, habits and expectations exceed the prepared scenarios. Recorded logs and Élise's notes preserve what the protocol cannot neatly classify.