The Visual Authority Trap

The challenge: Students tend to perceive attractive looking results as more trustworthy.

This is the aesthetic bias, a behavioral phenomenon where humans instinctively attribute high credibility and truthfulness to objects that are visually pleasing. When a generative AI tool gives a perfect looking presentation or paper, the impulse is to perceive it as authoritative.

Solution 

We trigger this impulse in class and we invite the students to reflect.

I created a visually appealing slide deck with the aid of an AI tool. The presentation has several authority cues, such as high-resolution, aesthetically pleasing and consistent imagery, clean layout and academic citations. I removed the AI watermarks. In class I share this presentation and ask the students to rate it on a scale from 0 to 100 and provide reasons for their scores. Once everybody has submitted their individual score, I reveal mine: 10/100. Every time I do this exercise my students are puzzled. Why? They all have given between 60 and 100 points to the presentation.

So I start to unpack the process of going beyond aesthetics. I open the presentation and, while commenting on its professional appearance, I scroll to the reference list. I open each reference one by one. As ‘404 Not Found’ pages appear, the classroom gets very quiet. I then prompt the students to discuss the implications of fake links. Are the titles genuine? Is the information cited in the presentation accurate? Which information is real and which is fabricated? How much is real and how much is false?

I then reassure them that this type of bias is more common than we realize. The aesthetics bias, which states that a product is perceived as easier to use when visually attractive, is a component of a broader cognitive bias known as the halo effect. The halo effect indicates that a single positive trait (such as physical beauty) influences the overall perception of a person or object leading to the assumption of other unrelated positive attributes. 

Example:

• We notice one standout positive quality: “He is well-dressed.”
• Halo effect: “Because he is well-dressed, we subconsciously assume he is also intelligent, trustworthy and capable.”. Our brain automatically fills in the blanks for other traits we haven’t actually observed yet.

In our case:

• The presentation uses a clean layout, consistent, pleasing imagery and citations.
• Halo effect: “This looks exquisite and organized, so the researcher must be an expert and the data must be flawless.”

In the context of AI, the halo effect is particularly dangerous because AI is exceptionally good in first impressions. When an AI produces a report or a presentation that looks professional, we are prone to subconsciously skip the critical evaluation of the actual content. High-quality design, perfect grammar and professional tone translates under the halo effect to “This looks like a high-level document, so the data must be verified and the logic must be sound.” In reality the AI might have “hallucinated” the information and the visual polish makes us less likely to fact-check.

Furthermore, large language models (LLMs), the engines of generative AI tools, are trained to be helpful and assertive. They rarely use phrases like “I guess”, “I’m not sure”, “I didn’t find data.” As a result AI tools can describe a non-existent historical event with the same authority the use to describe laws of physics. This absolute certainty in the writing style further accentuates our halo perception: The AI sounds so confident and articulate that we instantly think it must be an expert on the topic, not considering that confidence is a linguistic style, not a measure of accuracy.

The last step of the exercise is to ask the students to find one claim in the deck that sounds plausible and try to find a real, working source that either confirms or denies it.

We want students to pause and question information’s accuracy, much like recognizing phishing attacks in everyday digital communications and introducing friction before sharing a post on social media to curb disinformation. Ultimately, the goal is to cultivate a habit of pausing and developing a healthy skepticism towards information, especially when polished.