Skip to content

Quiz: Responsible and Ethical Use of AI

Test your understanding of accountability, disclosure, data governance, intellectual property, bias, over-reliance, and lab AI policies with these 10 review questions.

1. What is the central principle of responsible AI use in research?

  1. Accountability
  2. Speed
  3. Automation
  4. Cost reduction
Show Answer

The correct answer is A. The researcher who uses the tool remains accountable for what is published or acted on under their name. Speed, automation, and cost reduction may be benefits, but they do not remove accountability.

Concept Tested: Responsible AI Use

See: ./index.md#responsibility-accountability-and-judgment

2. What is the common position of major publishers regarding listing an AI system as an author?

  1. AI can be a co-author if it contributed substantially
  2. AI can be listed only in the acknowledgments
  3. AI authorship is allowed with disclosure
  4. AI systems are not listed as authors because they cannot be accountable
Show Answer

The correct answer is D. Major publishers hold that authorship requires accountability and the ability to approve the final version, which AI systems cannot provide. Substantive AI use is disclosed, but the AI is not an author.

Concept Tested: Authorship Policy

See: ./index.md#integrity-disclosure-and-authorship

3. Why should a log of AI use be kept during the work rather than reconstructed at submission time?

  1. Because disclosure is not required if the log is missing
  2. Because memory is unreliable and disclosure must be accurate
  3. Because the log is needed to train the model
  4. Because journals require daily logs
Show Answer

The correct answer is B. An accurate disclosure statement depends on recording tool names, versions, dates, and purposes as the work happens. Reconstructing it from memory invites omissions and errors. The log is not used to train models, and journals do not typically mandate daily logs.

Concept Tested: AI Disclosure

See: ./index.md#integrity-disclosure-and-authorship

4. Why are unpublished results a special concern when using hosted AI tools?

  1. Because unpublished results are always legally secret
  2. Because hosted tools cannot process research data
  3. Because submitting them to a hosted tool discloses them to a third party
  4. Because unpublished results are always public
Show Answer

The correct answer is C. Pasting unpublished measurements or designs into a hosted tool sends them to the provider under that provider's terms, which may allow retention or training use. They are not necessarily legally secret, but they are valuable and should be classified before sharing.

Concept Tested: Unpublished Results

See: ./index.md#privacy-confidentiality-and-governance

5. What is the main risk of over-reliance on AI tools in research?

  1. The researcher may stop being able to detect errors in the output
  2. The tools will become more expensive
  3. The tools will refuse to run
  4. The research will be completed too quickly
Show Answer

The correct answer is A. Over-reliance means accepting output without the checks it requires, which erodes the researcher's ability to spot errors. Cost, tool availability, and speed are not the central risks.

Concept Tested: Over-Reliance

See: ./index.md#dependence-and-skill

6. An agent substantially rewrote two paragraphs of the Discussion section in a submitted manuscript. Under the chapter's illustrative disclosure policy, what should be done?

  1. Nothing, because only text drafted entirely by AI needs disclosure
  2. Disclose the AI use in the paper
  3. List the AI as a co-author
  4. Record it only in the project log
Show Answer

The correct answer is B. The illustrative policy requires disclosure in the paper for text drafted or substantially rewritten by an AI tool that appears in a submitted manuscript. AI is never an author, and substantial rewriting is not merely grammar or formatting help.

Concept Tested: Citation of AI Use

See: ./index.md#integrity-disclosure-and-authorship

7. A researcher wants to use an AI tool on foundry process design kit model files covered by an NDA. Under the illustrative data classification, what is the appropriate action?

  1. Use any tool, because the files are technical
  2. Use an approved hosted tool after disclosure
  3. Use only a local model, if the agreement allows; otherwise no AI tool
  4. Use the tool but do not tell the foundry
Show Answer

The correct answer is C. Vendor files under NDA are in the most restrictive class. They may be given only to a local model if the agreement permits; otherwise no AI tool should process them. Using them covertly violates the agreement.

Concept Tested: Confidential Data

See: ./index.md#privacy-confidentiality-and-governance

8. A student wants to keep their own skills sharp while using AI. Which practice best counters skill atrophy?

  1. Let the agent do all first drafts and only review them
  2. Use the strongest model for every task
  3. Never explain AI-assisted work to others
  4. Do the first derivation or first draft of a core argument yourself
Show Answer

The correct answer is D. Doing core work yourself before using the tool preserves the reasoning skills that expertise is built from. Delegating first drafts to the strongest model and avoiding explanation accelerate atrophy.

Concept Tested: Skill Atrophy

See: ./index.md#dependence-and-skill

9. A lab policy states that authors are accountable for all content and that AI is never an author. What does this rule imply about AI-assisted content?

  1. Substantive AI use must be disclosed because humans take responsibility for the final content
  2. AI use does not need to be disclosed if a human reviewed it
  3. Disclosure is optional because accountability is already covered
  4. Only AI-generated figures need disclosure
Show Answer

The correct answer is A. Accountability means the human author stands behind the work, so readers and reviewers need to know what was machine-assisted. Review alone does not remove the need for disclosure, and the policy is not limited to figures.

Concept Tested: Accountability

See: ./index.md#responsibility-accountability-and-judgment

10. A researcher notices that AI-assisted literature surveys keep returning papers from the same set of high-profile venues. What is the most appropriate response?

  1. Accept the results because those venues are the most reliable
  2. Stop using AI for literature surveys
  3. Assume the other venues have no relevant work
  4. Recognize output bias and explicitly search across other venues, regions, and languages
Show Answer

The correct answer is D. Output bias can favor well-represented venues, languages, and recent work. The countermeasure is to deliberately vary inputs and check what is missing, not to assume the returned sample is complete or to abandon the tool.

Concept Tested: Bias and Fairness

See: ./index.md#fairness-environment-and-dual-use