Personal Agent Comparison Criteria Matcher¶
Run the Personal Agent Comparison Criteria Matcher MicroSim Fullscreen
Edit in the p5.js Editor
About This MicroSim¶
Personal agents change from month to month, but the questions a researcher should ask of them do not. This MicroSim practises turning a worry about an agent into the comparison criterion that addresses it. Each of eight questions belongs under exactly one of eight criteria:
| Criterion | What it covers |
|---|---|
| Where it runs | Where the computation and the data are located: your laptop, the vendor's cloud, or a server you manage |
| Blast radius | How far a failure, compromise, or injected instruction can spread |
| Approval model | The policy that decides which actions need a human |
| Usage metering | How consumption is counted, capped, and billed |
| Data exposure | What happens to the data given to the agent: training, retention, access |
| Memory | Whether you can inspect and control what the agent keeps between sessions |
| Claim verification | Independent evidence behind what the vendor advertises |
| Activity audit | The record of the agent's actions that is available for review |
The criteria come from the section "Risks That Define the Criteria" in Chapter 19. The questions are invented for illustration and do not describe any one product. The order of the criterion buttons is shuffled for each learner, so the position of a button tells you nothing about the answer.
How to Use¶
- Read the question in the dark-blue box. The question on screen is Which criterion does this question test?
- Press one of the eight criterion buttons below the drawing area. Your choice is marked with a filled circle and a dark background. You can change it until you submit.
- Press Submit. Each question gets one attempt. The feedback reads Correct: criterion. (with a check mark) or This asks about criterion. (with an X), followed by the "Why" sentence. The correct criterion's button is marked with a check mark and a solid border; a wrong pick is marked with an X and a dashed border.
- Press Next Question to continue. The questions come in a fixed order.
- After question 8, press See Score. The end screen shows n of 8 correct, whether you met the mastery threshold of 7, and a one-line test for each criterion that you could run on a real product. On narrow screens, press More Tests to page through the tests.
All controls are buttons that can be reached with the Tab key and pressed with Enter or Space. To try again with a new criteria order, reload the page.
Lesson Plan¶
Learning objective: The learner will classify each of eight questions about a personal agent under the comparison criterion it addresses, with at least 7 of 8 correct.
Bloom level and verb: Understand (L2), classify.
Evidence of mastery: For each of eight questions the learner commits one of eight criteria. A classification is correct when it equals the answer key. Mastery is at least 7 of 8 correct on the first attempt. Reading the list of criteria is exploration and is not evidence.
Misconceptions to watch for:
- A comparison is about features and price only.
- A vendor's statement about safety is itself the verification.
- Where an agent runs is the same as what it can reach.
Duration: 10 minutes.
Prerequisites: agent blast radius, usage metering, agent data exposure, vendor claim verification, overnight activity audit, cloud agent computer, self-hosted agent, and the approval model, as defined in the sections "What Makes an Agent Personal and Always-On" and "Risks That Define the Criteria" of Chapter 19.
Follow-up: Pick one agent you might use and run two of the one-line tests from the end screen with a throwaway account and synthetic data. Record what you found for each criterion.
References¶
- Chapter 19: Always-On Personal Agents for Research - the section "Risks That Define the Criteria" defines the eight criteria used in this MicroSim.