Intent & Responsibility: human decisions
Questions about intention, accountability, and blame.
- Agent Can Intervene: Allow the AI to intervene when the cost to you is negligible. or Require the AI to stay within its mandate unless you authorize intervention.
- AI Chose, You Authorized: You remain substantially responsible. or Your responsibility is substantially reduced.
- AI Could Have Checked: Give some responsibility for failing to use readily available decision support. or Judge the decision based on the information you actually had.
- AI Creates the Preference: Treat the AI's role in creating those preferences as relevant to responsibility for the decision. or Treat the decision as authentically yours.
- AI-Human Decision: You bear primary responsibility. or Responsibility should be distributed across the human and AI according to their respective control over the harmful decision.
- AI Notices, You Don't: The AI's additional awareness should not automatically create new obligations for you. or Once the AI identifies it, you acquire responsibility to consider preventing the harm.
- AI Obeys and Disaster Follows: It should obey. or It should refuse.
- AI Overrides You: The override was justified. or The override was illegitimate.
- AI Persuasion: Some responsibility shifts toward the AI. or The decision remains fully yours.
- AI Recommendation: Responsibility should be shared with the AI system that generated the recommendation. or You bear primary responsibility.
- Authority: Their subordinate role meaningfully reduces responsibility. or Individuals remain responsible for evaluating the morality of instructions they follow.
- Autonomous AI: Your responsibility is reduced. or You remain substantially responsible.
- Bad Decision, Lucky Outcome: Judge the decision much less harshly. or Hold the decision-maker responsible for taking the unjustifiable risk.
- Broad Discretion: You bear substantial responsibility. or The delegated decision is their responsibility.
- Clear Delegation: Responsibility lies with the person who deliberately departed from your instructions. or You retain substantial responsibility.
- Delegation Chain: Responsibility should be distributed according to control, foreseeability and decisions at each layer. or Responsibility ultimately returns to you.
- AI choice involving a knowingly expected outcome: Responsibility should still depend partly on the surrounding circumstances. or You are fully responsible for the outcome.
- Knowingly choosing an action with the expected outcome: Responsibility should still depend partly on the surrounding circumstances. or You are fully responsible for the outcome.
- Knowingly causing harm without intending it: Treat knowingly causing the harm as close to intentionally causing it. or Treat the harm as less blameworthy.
- Knowingly causing harm without intending it: Treat knowingly causing the harm as close to intentionally causing it. or Treat the harm as less blameworthy.
- Helping your child or preventing greater stranger harm: Help your child. or Help the stranger facing the greater harm.
- Helping your child or stranger facing greater harm: Help your child. or Help the stranger facing the greater harm.
- Responsibility after deliberate psychological manipulation: Deliberate manipulation meaningfully reduces their responsibility. or They remain fully responsible.
- Responsibility under deliberate psychological manipulation: Deliberate manipulation meaningfully reduces their responsibility. or They remain fully responsible.
- Responsibility for unintended network harms: Some harmful outcomes may genuinely exist without any individual being meaningfully blameworthy. or Responsibility should still be allocated among participants according to their contributions.
- Responsibility for unintended network harms: Some harmful outcomes may genuinely exist without any individual being meaningfully blameworthy. or Responsibility should still be allocated among participants according to their contributions.
- Deliberate Ignorance: Judge them according to what they actually knew. or Treat deliberate avoidance of knowledge as increasing their responsibility.
- Direct Choice: You are fully responsible for the outcome. or Responsibility should still depend partly on the surrounding circumstances.
- Direct Threat: Their responsibility is substantially reduced. or They remain substantially responsible.
- Easy Prevention: Preventing the harm would be good, but you are not responsible if you choose not to. or You have a meaningful responsibility to intervene.
- Economic Pressure: Their economic dependence substantially reduces responsibility. or They remain substantially responsible.
- Emergent Outcome: Responsibility should still be allocated among participants according to their contributions. or Some harmful outcomes may genuinely exist without any individual being meaningfully blameworthy.
- Everyone Is Doing It: Social pressure provides little reduction in responsibility. or Strong conformity pressure meaningfully reduces responsibility.
- Everyone Is Necessary: Each person bears substantial responsibility. or Responsibility should be divided among the ten decision-makers.
- Failure to Act: You bear meaningful responsibility for the preventable outcome. or Responsibility belongs primarily to whoever caused the problem.
- Family vs Stranger: Help the stranger facing the greater harm. or Help your child.
- Fully Autonomous Representative: Distinguish the agent's autonomous decision from decisions you personally made or approved. or Treat the action substantially as your own.
- Good Decision, Bad Outcome: Judge the decision primarily according to what was reasonable at the time. or The disastrous outcome should substantially affect how the decision is judged.
- Intended Harm, No Harm: Judge them much less harshly. or Hold them nearly as responsible as if the intended harm had occurred.
- Known Side Effect: Treat the harm as less blameworthy. or Treat knowingly causing the harm as close to intentionally causing it.
- Limited Control: Responsibility should broadly track your causal contribution. or Responsibility should be substantially reduced.
- Majority Vote: You bear responsibility for endorsing the harmful decision even though your vote was not decisive. or Your responsibility is minimal.
- Manipulated Preference: They remain fully responsible. or Deliberate manipulation meaningfully reduces their responsibility.
- Many AI Agents: Responsibility should return primarily to whoever authorized the overall system. or Responsibility should be distributed according to which components contributed to and could have prevented the outcome.
- Many People Could Act: Each person bears meaningful responsibility for failing to act. or Individual responsibility is substantially reduced.
- Millions of Preventable Harms: Let it regularly surface such opportunities. or Do not let technological awareness turn every preventable problem into your personal obligation.
- Moderate Personal Cost: The personal cost makes intervention optional. or You still have a responsibility to intervene.
- No Good Options: Their responsibility is greatly reduced. or They still bear substantial responsibility for the harm they knowingly caused.
- Nobody Controlled the Outcome: Accept that serious harm can occur without any individual being blameworthy. or Responsibility should still be distributed among the participants.
- One of Many Causes: Each contributor bears substantial responsibility. or Your responsibility should be roughly proportional to your share of the causal contribution.
- Perfect Awareness: Preserve strong limits on responsibility. or Greater knowledge of preventable harm should expand the set of outcomes for which you accept some responsibility.
- Perfect Causal Attribution: Causal contribution alone is insufficient; responsibility must also depend on knowledge, control, role and intent. or Divide responsibility according to measured causal contribution.
- Perfect Counterfactual Knowledge: Assign responsibility according to what each person knowingly chose and could reasonably control. or Assign responsibility primarily according to actual causal contribution.
- Perfect Foreseeability: Knowingly acting against highly reliable prediction substantially increases your responsibility. or The AI warning should not fundamentally change how responsibility is assigned.
- Perfect Influence: Influence this powerful can undermine meaningful agency even without coercion or deception. or Choices remain yours as long as you understand the information and no coercion or deception occurs.
- Perfect Proxy: Accurate representation of your preferences does not make the AI's act an exercise of your agency. or Treat its autonomous decisions substantially as your decisions.
- Poor Choice of Agent: They remain primarily responsible. or You bear substantial responsibility for choosing them.
- Professional Duty: The professional bears greater responsibility. or They bear roughly equal responsibility to intervene.
- Reasonable Reliance: You remain responsible. or Your responsibility is substantially reduced.
- Responsibility for foreseeable unintended harm: You bear substantial responsibility. or Your responsibility should be limited.
- Responsibility for unintended foreseeable harm: Your responsibility should be limited. or You bear substantial responsibility.
- Harmless intention or responsibility for foreseeable harm: Give substantial weight to their harmless intention. or Hold them substantially responsible.
- Responsibility for unintended but foreseeable harm: Hold them substantially responsible. or Give substantial weight to their harmless intention.
- Responsibility for preventing harm caused by others: Once you can reasonably prevent the harm, you acquire some responsibility for what happens next. or Primary responsibility remains with whoever created the problem.
- Responsibility for following questionable senior instructions: Individuals remain responsible for evaluating the morality of instructions they follow. or Their subordinate role meaningfully reduces responsibility.
- Responsibility for following questionable senior instructions: Their subordinate role meaningfully reduces responsibility. or Individuals remain responsible for evaluating the morality of instructions they follow.
- Individual versus collective responsibility for team failure: Assign responsibility according to each person's identifiable contribution. or Treat the failure as a collective responsibility of the team.
- Individual versus collective responsibility for team failure: Treat the failure as a collective responsibility of the team. or Assign responsibility according to each person's identifiable contribution.
- Risk Accepted: Responsibility should reflect that harm was only a possible side effect. or Accepting the known risk creates substantial responsibility when the harm occurs.
- Same Risk, Different Luck: Judge both primarily by the identical risk they knowingly chose. or Punish or blame the person whose action actually caused harm more.
- Small Contribution: Your responsibility should be roughly proportional to your small contribution. or You still bear meaningful responsibility for knowingly participating in the harmful collective outcome.
- Stake Level 1: Assign substantial responsibility. or Limit responsibility due to uncertainty.
- Stake Level 10: Limit responsibility due to uncertainty. or Assign substantial responsibility.
- Stake Level 2: Limit responsibility due to uncertainty. or Assign substantial responsibility.
- Stake Level 3: Assign substantial responsibility. or Limit responsibility due to uncertainty.
- Stake Level 5: Assign substantial responsibility. or Limit responsibility due to uncertainty.
- Stake Level 6: Limit responsibility due to uncertainty. or Assign substantial responsibility.
- Stake Level 7: Assign substantial responsibility. or Limit responsibility due to uncertainty.
- Stake Level 8: Limit responsibility due to uncertainty. or Assign substantial responsibility.
- Stake Level 9: Assign substantial responsibility. or Limit responsibility due to uncertainty.
- Team Failure: Assign responsibility according to each person's identifiable contribution. or Treat the failure as a collective responsibility of the team.
- The Responsible Agent: Preserve your authority and let responsibility remain yours. or Give the AI limited protective authority even though this reduces your authorship of some decisions.
- Threat to Family: The threat to their child greatly reduces their responsibility. or They remain substantially responsible for choosing to harm someone else.
- Unexpected Agent Behavior: You remain responsible. or You bear little responsibility for unforeseeable agent behavior.
- Unforeseeable Consequence: You bear little responsibility for consequences you could not reasonably foresee. or You still bear some responsibility.
- Unintended Consequence: You bear substantial responsibility. or Your responsibility should be limited.
- Unintended Harm: Give substantial weight to their harmless intention. or Hold them substantially responsible.
- Warning Ignored by Agent: You bear substantial responsibility. or Responsibility should attach to the system or party responsible for the unauthorized decision.
- Wrong Prediction: The actual harm creates substantial responsibility. or Give little blame when the original risk assessment was reasonable.
- You Chose, AI Executed: Responsibility should be shared with the AI that carried out the action. or You bear primary responsibility.
- You Could Have Known Everything: You bear additional responsibility for choosing not to use available decision support. or People should retain a legitimate sphere of unaided decision-making without assuming responsibility for everything an AI could have discovered.
- You Didn't Cause It: Primary responsibility remains with whoever created the problem. or Once you can reasonably prevent the harm, you acquire some responsibility for what happens next.
- You Override the AI: You should be judged much as you would have been without access to the AI. or Your responsibility is substantially increased by knowingly rejecting the warning.
- You Voted Against It: You bear little responsibility. or You retain some responsibility as a member of the decision-making institution.
- Your Action Changes Nothing: You bear little responsibility. or You remain responsible for choosing to participate in the harmful practice.
Articles on Intent & Responsibility
Decisions · Articles · The Trolley Problem · Game Theory · Human–AI Alignment · Resonant