Response latency and AI silence
The time an AI does not answer — what fills a response delay
When an AI does not answer straight away, what does a person fill the gap with?
Response latency is a performance number measured in seconds, but to the person waiting it is an event that has to be interpreted. Silence is not empty — a person fills it with a reason, and that reason changes what they trust and do next.
The study on this topic is under submission and in review. No findings or conclusions are published yet.
Why this question
Conversational AI is usually judged on the content of its answers. But in real use, the thing a person meets most often is not the answer — it is waiting for one. The wait goes undesigned, and what a person thinks during it never reaches the log.
An explanation moves into the empty seat: it is broken, it is ignoring me, it is thinking because the question is hard, I must have pressed something wrong. Which explanation arrives decides whether the same three seconds are one experience or another.
The work in progress
A study on how adults with ADHD endure and interpret silence from an AI is under submission. It is in review, so no findings or conclusions can be stated yet, and the title and scope may still change. What is written here goes only as far as the question being asked.
That sensitivity to waiting differs between people bears directly on design. A delay tuned to the average user is nothing to one person and a reason to leave for another.