The AI Taxonomy.
Terms  /  Human-Agent Relationship  /  Confidence Decay
07 · Human-Agent Relationship

Confidence Decay

The gradual erosion of user trust as a system repeatedly loses context, hallucinates, or fails to follow through.

Confidence Decay is the emotional counterpart to system reliability. Every time the agent forgets something the user already explained, every time it claims to have done something it didn't, every time it produces subtly wrong output that wastes the user's time — trust erodes. Unlike a single dramatic failure (which is obvious and can be addressed), Confidence Decay is cumulative and gradual. The user doesn't lose trust all at once; they slowly develop a background anxiety about whether the agent's output can be relied upon. Eventually, they spend more time verifying than the agent saves them.

Example
A developer's experience over six weeks: Week 1: agent is helpful and accurate, developer trusts output. Week 2: agent hallucinates a function signature, costing 30 minutes of debugging. Week 3: agent forgets a decision from the previous session, requiring re-explanation. Week 4: agent introduces a subtle bug that makes it to staging. Week 5: developer starts copy-pasting every code suggestion into a separate verification step. Week 6: developer manually reviews every line the agent writes, effectively doubling their workload. The agent is still producing useful output, but Confidence Decay has eliminated the time savings.

More in Human-Agent Relationship

Human-Code Divergence

The gap created between a user and their codebase when an agent mediates all development.

Cognitive Load

The mental burden placed on the human when working with AI systems.

Agency Erosion

Gradual loss of human decision-making authority as agents take over more workflow without checkpoints.

Skill Atrophy

Decay of human technical skills from prolonged delegation to agents.

Automation Dependency

Reliance on agents to the point where the human can't perform the task without them.

Trust Without Verification

Accepting agent output without review — a precondition for Execution Hallucination going undetected.

Capability Illusion

The user believes the AI can do more than it actually can, leading to over-delegation and undetected failures.

Learned Helplessness

The user stops attempting tasks they could do because the agent is "supposed to handle it."