“I don’t know anything about laptops!” – User Perception of Digital Product Advisors Adapting to Their Knowledge Levels

This study investigates how digital product advisors can effectively communicate complex technical information to users with varying levels of domain knowledge. The authors focused on the "knowledge-level adaptation" challenge, where systems must balance the needs of novices—who often struggle with technical jargon—and experts—who may find simplified explanations redundant or patronizing.

To test this, the researchers developed "Cleo," a rule-based chatbot that guides users through a laptop search. They conducted a between-subjects experiment with 251 participants, assigning them to one of four conditions:

• T (Baseline): Technical specifications only.
• TC: Technical specifications plus performance categories (e.g., "mid-range").
• TE: Technical specifications plus plain-language attribute explanations.
• TCE: A combination of technical specifications, performance categories, and attribute explanations.

The study reveals that novices derive significant value from supplementary information. Specifically, conditions that included attribute explanations (TE and TCE) were rated as more helpful and led to higher perceived learning than the baseline. Furthermore, novices found the combined approach (TCE) to be the most appropriate in terms of information quantity, suggesting that performance categories are most effective when paired with explanations that clarify the underlying technical concepts.

Crucially, the study found no significant differences in how experts perceived the various conditions. This suggests that providing extra, novice-friendly information does not "alienate" experts or detract from their interaction, supporting the feasibility of a single, inclusive interface design.

Based on these results, the authors propose four design guidelines for conversational commerce:

1. Use TCE by default: Provide technical specs, categories, and explanations to ensure all users have the context they need.
2. Keep a single inclusive interface: Avoid creating separate "novice" and "expert" modes, as the inclusive design works for both groups.
3. Avoid standalone categories: Performance labels (like "high-end") are ambiguous without accompanying explanations of why they are classified that way.
4. Support user agency: Allow users to personalize the interaction based on their stated use case (e.g., gaming vs. word processing).