How People Delegate AI Actions — and Assign Responsibility in Digital Life — 2026 Data
This page summarises findings from the Human Clarity Institute’s Delegated Action & Responsibility 2026 dataset, based on 344 valid responses across six English-speaking countries.
The research examines when people delegate actions to AI systems, how closely they monitor outcomes, where they intervene, and how responsibility is assigned when AI-supported decisions go wrong.
Within the broader Human–AI decision system, this dataset focuses on how decision tasks are distributed between people and AI, including delegation, monitoring, intervention, and responsibility for outcomes.
See the full system model:
A data-driven model of how AI reshapes thinking, confidence, control, and decision behaviour
View the Delegated Action & Responsibility 2026 Dataset
What the data shows
People tend to delegate actions to AI selectively, most often when decisions feel complex, uncertain, or time-pressured. In these situations, delegation is used to support progress, while people continue to stay involved in how outcomes unfold.
Four behavioural signals stand out in this dataset: comfort with delegating actions to AI is mixed; delegation is more likely under conditions of complexity or uncertainty; monitoring behaviour remains extremely high; and responsibility for outcomes is often shared rather than transferred.
Together, these findings show that delegation occurs within a system of ongoing human oversight and accountability. Across HCI’s wider behavioural datasets, this pattern increasingly suggests that people often attempt to preserve agency and responsibility even when parts of decision-making are delegated to AI systems.
People often allow AI systems to act on their behalf, but typically remain actively involved by monitoring outputs, stepping in when needed, and retaining responsibility for the final outcome. In practice, delegation appears to function less as a transfer of authority and more as a system of supported action combined with ongoing human oversight.
Delegation is not evenly applied across all decisions. It tends to increase when decisions require effort, speed, or additional support, rather than occurring as a default behaviour.
Feel comfortable delegating AI actions
Comfort with handing actions to AI is present, but not dominant.
People are more likely to feel comfortable delegating actions when the task feels lower-risk or familiar, rather than across all situations.
Pay close attention to AI outcomes
Monitoring behaviour is widespread.
People typically monitor what AI systems do after delegating actions, rather than leaving outcomes unchecked.
See responsibility for negative outcomes as shared
Responsibility is most often distributed between the user, the system, and the organisation.
People often assign responsibility across multiple parties when outcomes go wrong, rather than attributing it solely to the AI system.
Intervene when decisions feel important
Intervention thresholds favour human involvement.
People are more likely to step in when decisions feel important or uncertain, rather than allowing systems to act without oversight.
Taken together, these findings suggest that delegation to AI is conditional rather than automatic, increasing in situations where decisions feel complex, uncertain, or require faster resolution. Even when people allow AI systems to act, they continue to monitor outcomes, intervene when necessary, and retain a sense of shared responsibility for the results.
In practice, people combine delegation with monitoring and selective intervention, using AI to support decisions under specific conditions rather than handing over control entirely.
By the numbers (from HCI data)
Prefer to make the decision themselves regardless of AI
A substantial share default to retaining full control rather than delegating decisions to AI.
Intervene when AI reasoning feels unclear
Many people step in when system outputs lack clarity.
People are more likely to intervene when they do not understand how a decision was made.
Of those comfortable delegating AI actions, most still closely monitor outcomes
Delegation is typically accompanied by active oversight.
Of those comfortable delegating AI actions, most still feel personally responsible
Responsibility remains with the individual even when actions are delegated.
Several behavioural patterns discussed on this page — including cognitive strain, behavioural reliance, delegated judgement, verification behaviour, and decisional uncertainty — were already documented within pre-generative-AI research on human decision-making. View the historical baseline
Patterns observed in the data
Delegation does not replace oversight
Even among those who report comfort delegating AI actions, monitoring behaviour remains extremely high.
Intervention is selective, not constant
People tend to intervene when decisions feel important or unclear, rather than intervening on every decision. This suggests that many people experience agency less through constant control and more through confidence that intervention remains possible when needed.
Responsibility remains distributed
Responsibility is often shared between the user, the system, and the organisation. Even so, responsibility rarely appears fully transferred away from the individual, particularly where people continue to monitor and intervene in outcomes.
Full delegation is rare
Very few people allow AI systems to act without intervention or oversight.
In practice, people tend to combine delegation with monitoring and selective intervention, reflecting how AI-supported decisions remain embedded within human oversight and responsibility.
Questions this data can answer
Are people comfortable delegating actions to AI?
42% report high comfort, while many remain cautious depending on the situation.
Do people monitor AI outcomes after delegation?
85% say they pay close attention to outcomes, showing that delegation is usually accompanied by oversight. Monitoring behaviour appears to function as a stabilising mechanism within AI-assisted decision-making, helping preserve intervention capacity even when actions are delegated.
Do people fully trust AI to act without intervention?
Only a small share accept actions without intervention, indicating that full delegation is rare.
When do people intervene in AI decisions?
People are more likely to intervene when decisions feel important or when system reasoning is unclear.
Who is responsible when AI-supported decisions go wrong?
38% assign responsibility as shared, reflecting distributed accountability rather than full transfer to AI.
Methodology
This dataset forms part of the Human Clarity Institute’s Human–AI Experience research programme, examining how people delegate actions to AI systems, how closely they monitor outcomes, where they intervene, and how responsibility is assigned when AI-mediated decisions produce negative outcomes. The study uses a cross-sectional online survey design and focuses on descriptive patterns in delegated action behaviour, monitoring and oversight, intervention thresholds, and responsibility attribution in digitally mediated life.
Data were collected via the Prolific research platform from adults across six English-speaking countries. Participants provided explicit consent for anonymised open publication as part of HCI’s open research programme.
Sampling & participants
- Final n: 344
- Countries: United Kingdom, United States, Canada, Australia, New Zealand, Ireland
- Eligibility: Adults aged 18+ from six English-speaking countries
- Recruitment platform: Prolific
The resulting dataset should be interpreted as a non-probability convenience sample and is not intended to represent national populations.
The cleaned dataset, variable dictionary, and reuse terms are publicly available through the HCI dataset repository: Delegated Action & Responsibility 2026 Dataset →
Data integrity
All percentages reported on this page are calculated from valid responses in the cleaned dataset (n = 344). Percentages are rounded to the nearest whole number for readability. Unless otherwise stated, summary percentages combine respondents selecting 5–7 on the 7-point agreement scale (slightly agree, moderately agree, or strongly agree).
Where percentages refer to categorical response distributions (such as intervention thresholds or responsibility attribution), the wording on the page makes that explicit. Distributions reflect the share of respondents selecting each option.
Participant IDs, timestamps, and direct identifiers were removed before publication as part of the anonymisation process.
This dataset is exploratory and descriptive in nature. It does not support causal inference and results should be interpreted as observed patterns within the survey sample.
This dataset is released as open research to support transparent analysis of delegated action behaviour, monitoring and oversight, intervention thresholds, responsibility attribution, and the human experience of acting with AI systems.
Data use and reuse terms are outlined in our Data Use & Disclaimer.
Explore further analysis on Human Clarity Insights, or browse the full collection of HCI research reports.