💻 computer science

"I don't know anything about laptops!" - User Perception of Digital Product Advisors Adapting to Their Knowledge Levels

This study demonstrates that in conversational e-commerce, augmenting technical product information with explanations and performance categories significantly improves novice users' perceived helpfulness and learning without negatively affecting expert users, leading to design guidelines for inclusive, adaptive digital advisors.

Original authors: Kevin Schott, Andrea Papenmeier, Daniel Hienert, Dagmar Kern

Published 2026-08-07

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Kevin Schott, Andrea Papenmeier, Daniel Hienert, Dagmar Kern

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are walking into a massive, high-tech library to find a specific book. If you are a librarian, you know exactly which aisle to head to, what the Dewey Decimal System means, and how to scan the spine for the perfect edition. But if you are a regular visitor who has never stepped foot in a library before, that same system looks like a secret code. You might feel overwhelmed, unsure if you're looking for a "QHD" or a "RAM," and you might just grab the first book that looks shiny. This is the daily reality for many people trying to buy complex things online, like laptops.

Scientists who study how humans and computers talk to each other call this field "Human-Computer Interaction." They are trying to figure out how to make digital assistants feel less like robots reading a manual and more like helpful friends. Two big ideas guide them: "Communication Accommodation," which is just a fancy way of saying "speak your listener's language," and "Cognitive Load," which is the idea that our brains have a limited amount of space for new information. If you dump too much jargon on someone, their brain gets full and they stop listening. If you give them too little, they feel lost. The big question is: How do you design a digital salesperson who can talk to both the expert librarian and the confused visitor without annoying either of them?

That is exactly what Kevin Schott and his team at GESIS and the University of Twente wanted to find out. They set up a digital experiment with a chatbot named "Cleo" to help 251 real people shop for a new laptop. They wanted to see if changing how Cleo explained technical specs would make a difference for beginners versus experts. They tested four different ways of talking:

The "Tech-Only" Mode: Just the raw numbers and abbreviations (e.g., "8–16 GB RAM").
The "Tech + Category" Mode: Numbers plus a simple label like "mid-range" or "high-end."
The "Tech + Explanation" Mode: Numbers plus a plain-English sentence explaining what the part actually does (e.g., "RAM helps your laptop handle multiple tasks at once").
The "Super-Helper" Mode: Numbers, labels, and explanations all rolled into one.

The results were a bit like a magic trick. For the beginners (the people who said they knew little about laptops), the "Super-Helper" mode was the clear winner. When Cleo explained why a certain amount of RAM mattered and gave it a friendly label, the beginners felt like they were actually learning something. They found the advice more helpful, trusted the chatbot more, and felt the amount of information was just right. Interestingly, the beginners didn't just like the explanations; they needed them to make sense of the categories. Without the "what does this do?" part, the labels like "mid-range" were just as confusing as the raw numbers.

Here is the best part: The experts (the people who knew a lot about laptops) didn't mind the extra help at all. While the researchers worried that giving beginners extra explanations might annoy the experts or make the chatbot feel slow and patronizing, that didn't happen. The experts didn't rate the "Super-Helper" mode any worse than the bare-bones "Tech-Only" mode. They seemed to be able to quickly skim past the extra words or use them as a quick check, without feeling burdened.

So, what does this mean for the future of shopping? The study suggests that we don't need to build two different chatbots—one for nerds and one for newbies. Instead, the best approach is to build one inclusive chatbot that gives everyone the full package: the technical specs, the easy labels, and the plain-English explanations. It turns out that when you help the beginners understand the complex world of laptops, you don't lose the experts; you just give everyone a better, more confident shopping experience. The paper doesn't claim this is a perfect, solved problem for every situation, but it strongly suggests that for technical products, being a little bit more explanatory is the key to making everyone feel smart.

Technical Summary: "I don't know anything about laptops!" – User Perception of Digital Product Advisors Adapting to Their Knowledge Levels

Problem Statement
Conversational commerce relies on digital assistants to facilitate search and decision-making in e-commerce. A critical challenge in this domain is adapting the presentation of complex product information to users with varying levels of domain knowledge (novices vs. experts). While communication accommodation theory and recipient design suggest that systems should tailor utterances to a user's presumed knowledge, existing research has largely focused on lexical or personality alignment. The adaptation of information type and amount for technical products remains underexplored. Novices often struggle with functional product information, relying on third-party opinions, while experts may be negatively affected by redundant information (the expertise reversal effect). There is currently no solid empirical understanding of how text-based product advisors should proactively present complex information to accommodate both groups without alienating either.

Methodology
The authors conducted a between-subjects online user study ( $N=251$ ) involving a rule-based text chatbot named "Cleo" designed for laptop search.

Participants: Users were recruited via Prolific and stratified into "novices" (self-rated knowledge 1–4 on a 7-point scale) and "experts" (self-rated 5–7).
Experimental Conditions: Participants interacted with Cleo under one of four information presentation formats regarding five laptop attributes (display, RAM, storage, CPU, GPU):
1. T (Technical Only): Numerical specifications and standard abbreviations (Baseline).
2. TC (Technical + Categories): Specifications plus performance labels (e.g., "mid-range configuration").
3. TE (Technical + Explanations): Specifications plus plain-language explanations of attribute function and impact, replacing specialized terms.
4. TCE (Combined): Technical specs, performance categories, and attribute explanations.
Procedure: Users answered questions about their intended use case (basic vs. advanced), and Cleo provided attribute-level recommendations. No specific products were shown.
Measures: Post-interaction surveys assessed perceived appropriateness of information quantity, perceived learning, relevance, trust, and helpfulness of supplementary information. Qualitative data was gathered via open-ended questions.
Analysis: Statistical analysis utilized Kruskal-Wallis tests with Dunn's post-hoc tests (Benjamini-Hochberg corrected) for within-group comparisons and Mann-Whitney U tests for between-group (novice vs. expert) comparisons.

Key Results

Novice Experience:
- Learning and Helpfulness: Novices reported significantly higher perceived learning and helpfulness in conditions with attribute explanations (TE and TCE) compared to the baseline (T) and categories-only (TC).
- Appropriateness: Novices rated the combined approach (TCE) as significantly more appropriate in terms of information quantity than the baseline (T) and TC.
- Qualitative Insights: The primary concern for novices was "Attribute Confusion" (struggling with technical terms), which was most prevalent in T and TC conditions (60%) and reduced in TCE (34%).
Expert Experience:
- No Significant Differences: Experts showed no significant differences in perceived appropriateness, learning, relevance, trust, or helpfulness across any of the four conditions.
- Qualitative Insights: Experts' primary request was for "Missing Specs & Attributes" (e.g., brand, price, weight) rather than explanations of existing attributes. They did not find the supplementary information detrimental to their experience.
Comparative Findings:
- Providing supplementary information beneficial to novices (explanations and categories) did not negatively impact the experience of experts.
- Experts rated the "Categories only" (TC) condition as more helpful than novices did, whereas novices found explanations (TE/TCE) significantly more helpful than categories alone.

Key Contributions and Design Guidelines
The study distills these findings into four concrete design guidelines for inclusive text-based product advisors in technical domains:

Use TCE by Default: The combination of Technical information, Categories, and Explanations (TCE) is the most effective approach. It supports novices in understanding and benefiting from performance categories without alienating experts.
Keep a Single Inclusive Interface: Rather than dynamically switching interfaces based on inferred knowledge (which risks misclassification), a single interface containing all information types serves both groups effectively.
Avoid Standalone Categories: Presenting performance categories without attribute explanations (TC) is insufficient for novices, who require functional context to interpret abstract labels.
Support User Agency and Personalize to Use Case: Both novices and experts expressed a desire for more options (comparisons of attribute values) and personalization tailored to specific use cases, rather than a single deterministic recommendation.

Significance
The paper claims that its findings provide an empirically backed method for implementing "knowledge-level recipient design" in conversational commerce. By demonstrating that a single, information-rich interface (TCE) can simultaneously support novice learning and maintain expert satisfaction, the authors challenge the assumption that systems must choose between simplicity for novices and conciseness for experts. The results suggest that effective communication accommodation in technical domains can be achieved by proactively providing context (explanations) alongside categorization, thereby fostering shared understanding and informed decision-making across the knowledge spectrum.

Technical Summary: "I don't know anything about laptops!" – User Perception of Digital Product Advisors Adapting to Their Knowledge Levels

More like this