Affordances for Conversational Search

Source: https://queryunderstanding.com/affordances-for-conversational-search-3e7f1c2d9b5a Author: Daniel Tunkelang

Summary

Conversational search promises to handle ambiguity through dialogue — but it fails in practice because of an affordance gap: users don’t know what kinds of clarifying questions or reformulations the system can handle.

What Is an Affordance?

In UX, an affordance is a design element that communicates how it can be used. A door handle affords pulling; a flat plate affords pushing. Good affordances make valid actions obvious.

The Conversational Search Affordance Gap

Traditional search boxes have clear affordances: type keywords, get results. Conversational interfaces claim to understand natural language — but users don’t know:

  • What can I ask?
  • Will it understand follow-up questions?
  • What reformulations are valid?

This leads to over-trust (users phrase queries the system can’t handle) and under-use (users revert to keyword search out of caution).

Shopify Example: Clarifying Questions

Tunkelang cites Shopify’s search feature that asks structured clarifying questions:

  • “Are you looking for a product or a store?”
  • “Do you mean X or Y?” These are designed affordances — the system signals exactly what it can distinguish.

Design Principles for Conversational Affordances

  1. Surface the query space — show examples of what kinds of queries work
  2. Offer structured clarifications — multiple choice > open-ended follow-up
  3. Progressive disclosure — start with keyword search, escalate to dialogue only when needed
  4. Fail gracefully — when the system doesn’t understand, say so and suggest alternatives

Relationship to Autocomplete

Autocomplete is a form of affordance: it shows users what queries are in-vocabulary. Conversational systems need analogous mechanisms.

Key Concepts

  • Affordance — design element communicating valid interactions
  • Affordance gap — mismatch between system capability and user mental model
  • Clarifying questions — structured dialogue to resolve ambiguity
  • Progressive disclosure — escalate to richer interaction only when needed

People