HCI · LOCAL SESSION WORKSHEET

Record a participant session

Use this sheet while a real person tries the prototype or chat application. All fields start blank. Downloads stay on this computer; nothing is submitted online.

Before starting

Explain that the interface is being tested, ask permission for anonymous notes, and let the participant stop at any point. Use a code such as P01 instead of a name. Record what happens before interpreting why it happened.

The two forms export independently. Download each completed form before closing or reloading this page. Do not use placeholder answers as observations.

HW3 · Fit Check

Allow about 10 minutes. The participant must be someone other than the author. Open Fit Check in another tab. Read one task at a time and ask the participant to think aloud. Offer help only after recording where they became stuck.

LLM · Compare A and B

Allow about 10–15 minutes. A has no waiting message; B shows “Preparing a reply…” before the first response text. Both use the same local model. The question is whether the wait and system state feel clearer, not whether the indicator makes inference faster.

  1. Run 启动LLM试用.cmd from the task pack and enter the participant code. Odd codes use A then B; even codes use B then A.
  2. Wait until the model is ready. Read the three prompts below, one at a time. After each reply, have the participant estimate the time from Send to the first visible response text, choose satisfaction from 1 (very dissatisfied) to 7 (very satisfied), and save the rating in the application.
  3. After all three turns, close the application. The launcher starts a fresh process for the second condition, clearing conversation history and repeating the same loading procedure. Use the same prompts and order.
  4. Ask the final preference question after both conditions. Record “no noticeable difference” if that is the participant's judgement. The existing subsecond technical timings may make the indicator hard to notice. Do not add a delay to the test.
Three prompts to use in each condition
  1. Explain one useful purpose of an Undo button.
  2. Give one example of unclear feedback in a software menu.
  3. Why might a small button be difficult to use on a phone?

Record failures and interruptions as outcomes; leave their first-text estimates and ratings blank. The application saves measured timings and ratings in the participant's sessions folder. Keep those CSVs with this qualitative record. Actual response times are hidden during manual testing to avoid influencing the estimate.