Explanation

The surfacing of reasoning behind system behaviour, balancing transparency with cognitive overhead. Explanations clarify deterministic logic, human actions, and AI-driven decisions to foster trust.

Types of explanation

Explaining (deterministic) system logic

Rules, workflows, why certain options appear, validation errors, etc.

Examples

• Explaining why an action is unavailable
• Clarifying validation errors or required fields in forms
• Communicating the next steps or expected actor actions in workflows

Domain explanations

Connecting system behaviour to real-world domain concepts. In complex systems, the interface embeds domain knowledge—process logic, regulatory requirements, industry practices. Domain explanations surface this embedded knowledge.

Examples

• Explaining why a workflow exists (regulatory requirement, industry standard, organisational policy)
• Clarifying domain terminology embedded in field labels or options
• Connecting data structures to the real-world entities they represent
• Grounding validation rules in domain constraints, not arbitrary system rules

Domain explanations transform system constraints from arbitrary-feeling to meaningful. They support domain learnability—helping users understand the domain through the tool.

Social/Collaboration explanations

Actions taken by other actors, notifications about shared resources, edits, status changes, etc. Transparency regarding the actions of other people to reduce uncertainty in collaborative or multi-actor environments.

Examples

• Explaining edits or changes made by another actor
• Providing context about shared resources or history
• Clarifying reasons for notifications or status updates initiated by other people

AI explanations

How the system arrived at a prediction, recommendation, or generated output.

Examples

• Explaining recommendation or prediction logic (“Why am I seeing this?”)
• Clarifying the limitations or confidence levels of AI-generated outputs
• Communicating factors influencing automated decisions or outputs

Dynamic explanation

On-demand explanations for any selected content, supporting contextualised learning.

When users select content, explanations can address different aspects:

• Interface mechanics: how to use this element
• Domain significance: why it exists and what it represents
• Personal relevance: connecting to the user’s actual data and context
• World knowledge: bridging system concepts to broader domain understanding

Amount of detail

1. Bare minimum: Brief, direct explanation for common scenarios
2. Moderate detail: expanded explanations providing additional context, suitable for actors needing further clarity or reassurance.
3. Extended report: explanations offering full transparency, suitable for complex decisions, troubleshooting, or expert actors who require in-depth understanding.

Progressive disclosure can be used to assist actors in transitioning between different levels of detail.

Components

Level 1. Indicator

Global

Describes a set of elements, aiming to convey the broader context influencing multiple objects.

TODO: Example: item ranking

Local

Explains a single value or item.

TODO: Example: Predicted match of an item

Level 2. Simple explanation

Users can access the explanation popover by clicking the explanation indicator. This popover offers contextual information beyond a single value and lists the factors that contributed to the result.

TODO: Example: explain price through costs, margin, and discount

Natural language Explanations

TODO: Examples: “I’m not sure. But…”, ”… Plese keep in mind, this information is uncertain.”,

Level 3. Extended explanation

TODO: Example: a drawer with instructions or help text.

Resources

• Andy Matuschak (2025) How might we learn? – contextualised study and dynamic explanations grounded in user’s actual work
• Wikipedia / Explainable artificial intelligence
• Mayer, R. E., & Moreno, R. (2003). Nine ways to reduce cognitive load in multimedia learning. Educational Psychologist, 38(1), 43-52.