Agent Transparency — an AXD Institute resource on agentic experience design, agentic commerce, trust architecture, and human agent interaction. Founded by Tony Wood..
| Dimension | Traditional UX | Agentic Experience Design (AXD) |
|---|---|---|
| Primary material | Attention and affordance | Trust and delegation |
| User state | Present, navigating | Absent, delegating |
| Design output | Screens and interfaces | Outcomes and constraints |
| Temporal model | Session-based | Relationship-based |
| Success metric | Task completion | Trust calibration |
Making AI agents transparent requires structural design, not surface-level reporting. Start by building legibility into the agent's architecture: every decision must be traceable to a specific delegation scope, constraint set, and reasoning path. Implement multi-level explanation interfaces - a one-sentence summary for casual review, paragraph-level detail for active monitoring, and full reasoning traces for audit. Design confidence communication that honestly conveys what the agent knows and wh
Agent legibility is the degree to which a human can understand an AI agent's reasoning, decisions, and behaviour without requiring technical expertise. It matters for trust because trust in autonomous systems depends on comprehension - humans cannot appropriately trust what they cannot understand. The AXD Institute distinguishes legibility from mere transparency: transparency means making information available, while legibility means making that information understandable. An agent that dumps it
Designing reasoning traces requires capturing decision-relevant information at every branching point in the agent's execution. Record what options were considered, what constraints were applied, what trade-offs were evaluated, and why the final choice was selected. Structure traces as decision trees with explicit branching logic that can be serialised and replayed. Include confidence levels at each decision point - not just the final output but the agent's certainty at each intermediate step. De
When an AI agent makes a mistake, transparency is the primary mechanism for trust recovery. The AXD approach to trust recovery through transparency involves four steps. First, immediate disclosure: the agent must proactively report the failure, not wait for the human to discover it. Second, causal explanation: the agent must explain what went wrong, why its reasoning led to the incorrect outcome, and what it failed to anticipate. Third, corrective transparency: the agent must show what it will d
Agent transparency and autonomy exist in a calibrated relationship defined by the AXD Trust Calibration Model. New agent relationships start with maximum transparency and minimum autonomy - every decision is explained, and the human approves significant actions. As the agent demonstrates consistent, legible decision-making, transparency requirements decrease and autonomy increases. This is not a one-way progression: failures trigger transparency increases and autonomy reductions. The key insight
Build confidence communication patterns that convey agent certainty without false precision - Build structured decision reports aligned with Implement transparency metrics that measure whether humans actually understand agent behaviour -