The offer of a system-generated recommendation that the actor can accept, modify, or reject, balancing automation with agency.
Input assistance
Conversation
Clues for what’s possible and what to do next.
Form
Auto-completion and smart defaults reduce cognitive load and prevent errors.
Form element
Individual form fields can provide contextual suggestions as users interact with them.
Suggesting relationships
The system can identify and propose connections between data points that users might not have considered.
Proposing solutions
Intelligent recommendations for actions or workflows based on current context and user patterns.
Certainty
Suggestions should communicate their confidence level to help users make informed decisions.
Adjustment
Users should be able to refine or modify suggestions to better match their needs.
Limitations
Fixation risk
Providing explicit suggestions can constrain creative exploration through fixation effects:
- Users anchor to AI-generated examples rather than exploring alternatives
- Ideas homogenise as multiple users converge on similar AI-suggested approaches
- Diversity decreases even when users generate initial ideas before seeing suggestions
- Lower elaboration depth compared to question-driven approaches (65% minor elaboration, 25% no elaboration in suggestion-based dialogue vs 55% substantive elaboration in question-driven)
Research shows suggestion-based dialogue reduces both individual creativity and collective diversity (Maier et al. 2025). The effect occurs because concrete examples limit the solution space users consider, even when suggestions are high quality.
Mitigation strategies:
- Use question-driven scaffolding or ambiguous analogies that leave interpretive space for human creativity rather than concrete examples
- Delay suggestions until after initial ideation to reduce anchoring
- Present suggestions as inspiration rather than recommendations
- Provide diverse suggestion sets from different conceptual domains
- Consider whether the task benefits more from exploration (use questions) or optimisation (use suggestions)
See Conversation dialogue forms for alternative interaction patterns that preserve diversity and elaboration depth.
Dismission and override
Users must have clear paths to reject suggestions and provide alternative input.
Related components
Resources & references
- Maier, Schneider, Feuerriegel (2025) Partnering with Generative AI: Experimental evaluation of human-led and model-led interaction in human-AI co-creation
Consequences
- the actor's option space now contains a concrete proposal, which is the help and also the anchor: shown before they have formed their own idea, it narrows what they go on to consider, and the alternatives they never reached aren't visible to them as a loss
- accepting is one gesture and declining is a judgement, so the offer only stays an offer while dismissing it is cheaper than weighing it
- a suggestion arriving without its confidence reads as an assertion; the actor needs to know how firm the ground under it is to weigh it against what they already think
Related patterns
Precedes
- Prompt — can provide an initial prompt
- Cognitive forcing functions — the actor's option space now contains a concrete proposal, which is the help and also the anchor: shown before they have formed their own idea, it narrows what they go on to consider, and the alternatives they never reached aren't visible to them as a loss
Serves
- Assistance — assistance's option-narrowing step — providing likely next actions and narrowing the option space
Enacts
- Adaptability — suggestions might sharpen on the actor's accepted-or-rejected signal, fitting their patterns over time
- Formality — suggestions ease the step from informal intent to formal expression, formalising by recognition rather than recall
- Agency — the system proposes and the actor disposes; accept, modify, or reject keeps the final say with the user
- Temporality — visualising the future — a suggestion puts a possible next state in front of the actor before they act
Complements
- AI completion — AI completion is a suggestion mechanism; fixation risk and dismissal apply directly
- Autocomplete — system-generated recommendations more broadly
- Autofill — autofill as a form of implicit recommendation
- Collaboration — Suggestions support collaborative decision-making through AI-assisted discovery
- Dynamic hyperlinks — suggested actions and connections
- Expressed uncertainty — a proposal's weight depends on how firm the ground under it is
- Prompt scaffolding — starters and example prompts seed the scaffold's slots
- Searching — "did you mean?" and related patterns that guide query formulation
- Template
- Transparent reasoning — can emerge from the visible reasoning process.
Tangentially related
- Localization — locale mismatch prompts
Alternatives
- Conversation — dialogue forms that provide alternatives to suggestion-based interaction
Related
- Approval gate — accepting proposed content versus authorising an act — the same accept/reject core at different stakes
- Good defaults — A weaker form of default ("Did you mean X?") or a dynamic default
- Inline interface — recommendations the actor can accept, modify, or reject in the flow.
- Delegation — agent might suggest opportunities it believes will help
Preceded by
- Embedded intelligence — the output is often a suggestion the actor can accept, reject, or modify
- Generated content — output can inform subsequent suggestions or recommendations.
Instantiated by
- Assisted task completion — the completion spectrum, including AI completion (inline generation) and next-best action (workflow-level recommendations)
- Next-best action — suggestion is the broader pattern; next-best action specialises it to recommending workflow steps