Are people really that easy to fool?

Inspiration

Conversations about misinformation often begin with the same assumption: people see false information, believe it, and pass it on.

But Hugo Mercier’s work on human reasoning suggests a more complicated picture. People are not simply passive receivers of information. They evaluate sources, compare claims with what they already know, consider other people’s reactions, and sometimes engage with content precisely because they distrust it.

This raised a question that I found more interesting than simply detecting fake news:

Are people really as easy to fool as we assume?

Testing that question is difficult. Surveys can ask whether someone believes a headline, but they remove the headline from the environment in which people normally encounter news. Real social-media platforms provide a more natural experience, but researchers cannot fully control their algorithms, interfaces, or content order.

I built NewsLab to bridge that gap: a controlled news environment that still feels like browsing an ordinary feed.

What it does

NewsLab is a browser-based laboratory for studying how people respond to information.

Researchers can create different experimental conditions and control what participants see, including the content, order, format, source cues, reactions, and surrounding interface. Participants then browse the feed naturally by scrolling, opening articles, reacting to items, or moving past them.

The platform records these interactions alongside survey responses. This makes it possible to distinguish between several behaviors that are often treated as equivalent.

A person may click a false headline because they believe it, but they may also click because they are skeptical, curious, amused, or surprised. Likewise, ignoring a headline does not necessarily mean recognizing that it is false.

NewsLab therefore does not try to determine whether a piece of news is fake. It gives researchers a way to study the people encountering it.

How we built it

I designed NewsLab around two connected experiences.

The researcher side allows experiments to be created and configured. Researchers can define conditions, add news items, preview the participant experience, manage study settings, inspect sessions, and export the resulting data.

The participant side presents the consent flow, assigns an experimental condition, displays the news feed, records interactions, and guides the participant through any final survey or completion page.

Behind the interface, NewsLab preserves the exact feed assigned to each participant and records events such as impressions, scrolling, article opens, reactions, visibility changes, and dwell time. The resulting data can be exported at participant, item, event, and survey level.

I also added optional AI-assisted content variation. Researchers can generate candidate versions of an item, compare them with the original, edit them, and decide whether to include them. The model does not automatically decide what belongs in an experiment; final control remains with the researcher.

Challenges we ran into

The largest challenge was turning an abstract research question into a trustworthy experimental workflow.

It was not enough to build a page that looked like a news feed. The system also had to preserve what each participant actually experienced. Reloading the page could not silently restart the experiment. Retried network requests could not create duplicate events. Preview sessions could not contaminate real study data. Changes made by a researcher could not retroactively alter an existing participant’s feed.

Another challenge was deciding what behavior means.

Clicks, reactions, and dwell time are observable, but belief is not. I had to avoid designing the platform around the assumption that engagement automatically indicates persuasion. Instead, NewsLab records multiple forms of behavior and leaves their interpretation to the experimental design.

Finally, publishing a public demonstration required a balance between openness and control. Visitors needed to explore the project without being able to damage the showcase experiment, while participant-facing routes still needed to remain interactive.

Accomplishments that we're proud of

The accomplishment I am most proud of is turning a theoretical disagreement into a working experiment.

NewsLab is not only a visual mock-up. It supports the full path from creating an experimental condition to assigning a participant, recording their behavior, collecting their answers, and exporting the results.

I am also proud that the project does not begin by assuming that users are gullible. It treats participants as active interpreters whose clicks, doubts, curiosity, prior knowledge, and social reasoning all matter.

Instead of building another tool that labels information for people, NewsLab creates a space in which researchers can observe how people make judgments for themselves.

What we learned

Building NewsLab changed how I thought about both misinformation research and software design.

I learned that behavioral logging is not merely an analytics feature. In an experiment, it becomes part of the scientific instrument. Small implementation choices can determine whether the resulting data is interpretable.

I also learned that realistic interaction and experimental control must be designed together. A perfectly controlled survey may fail to capture natural behavior, while a realistic feed may produce data that cannot be explained if too many variables remain hidden.

Most importantly, I learned that attention, engagement, belief, and persuasion should not be treated as the same thing. A good experimental platform should preserve those distinctions rather than collapse them into a single metric.

What's next for NewsLab: Controlled News-Feed Experiments

The next step is to use NewsLab for concrete studies inspired by questions about epistemic vigilance, trust, source credibility, social signals, and misinformation.

Future versions could support richer experimental manipulations, longitudinal studies, collaborative researcher accounts, stronger accessibility, multilingual experiments, preregistered study templates, and more advanced analysis tools.

I would also like to work with researchers to test whether the platform captures the variables they actually need and to develop experiments that compare stated belief with observable behavior.

NewsLab began with a simple question:

Are people really that easy to fool?

Its goal is not to answer that question in advance. Its goal is to make it possible to test.