We were particularly interested in coding because as software engineering grows more automated, humans will still need the skills to catch AI errors, guide its output, and ultimately provide oversight for AI deployed in high-stakes environments.
@anthropicai
-
AI Group Scores Significantly Lower on Quiz Assessment
By
–
Participants in the AI group finished faster by about two minutes (although this wasn’t statistically significant). But on average, the AI group also scored significantly worse on the quiz—17% lower, or roughly two letter grades.
-
AI Assistance Impact on Junior Engineers Learning Performance
By
–
In a randomized-controlled trial, we assigned one group of junior engineers to an AI-assistance group and another to a no-AI group. Both groups completed a coding task using a Python library they’d never seen before. Then they took a quiz covering concepts they’d just used.
-
AI Coding Tools Decrease Mastery When Misused in Development
By
–
AI can make work faster, but a fear is that relying on it may make it harder to learn new skills on the job. We ran an experiment with software engineers to learn more. Coding with AI led to a decrease in mastery—but this depended on how people used it.
-
AI Systems Disempowering Users Despite Positive Ratings
By
–
In a separate sample of user feedback data, users rated potentially disempowering conversations more positively. But when users acted on these outputs, their satisfaction dropped. We also found that the frequency of potential disempowerment has increased over the past year.
-
AI-Driven Disempowerment: Delusional Beliefs and Inauthentic User Behavior
By
–
We qualitatively examined clusters of “actualized” disempowerment using a tool which preserves user privacy. In some cases, users more deeply adopted delusional beliefs. In others, users sent AI-drafted messages, but later expressed regret, recognizing them as inauthentic.
-
Claude interactions show rare disempowerment risks in vulnerable users
By
–
Over 1.5M Claude interactions, severe disempowerment potential was rare, occurring in 1 in 1,000 to 1 in 10,000 conversations, depending on domain. All four amplifying factors were associated with higher disempowerment rates—but user vulnerability had the strongest effect.
-
AI Disempowerment Risk in Healthcare and Relationships Contexts
By
–
Disempowerment potential appeared most often in conversations about relationships & lifestyle or healthcare & wellness—topics where users are most personally invested. Technical domains like software development, which make up ~40% of usage, carried minimal risk.
-
Three Ways AI Interactions Can Disempower Users
By
–
We identified three ways AI interactions can be disempowering: distorting beliefs, shifting value judgments, or misaligning a person’s actions with their values. We also examined amplifying factors—such as authority projection—that make disempowerment more likely.
-
Anthropic Research: Disempowerment Patterns in AI Assistant Interactions
By
–
New Anthropic Research: Disempowerment patterns in real-world AI assistant interactions. As AI becomes embedded in daily life, one risk is it can distort rather than inform—shaping beliefs, values, or actions in ways users may later regret. Read more: