A public AI trust study found that factual knowledge can change attitudes toward government automation more effectively than direct personal experience with the technology. Researchers Yotam Margalit and Shir Raviv studied how citizens judge machine-learning systems used in public administration. The findings were circulating on April 5, 2026, as governments debated how much explanation, transparency and safeguard detail they owe people affected by automated decisions.
Margalit and Raviv focused their research on the psychological divide between interaction and understanding. Participants engaged in a controlled experiment designed to replicate real-world scenarios where algorithms assist in resource allocation or regulatory enforcement. Exposure to these tools failed to produce a measurable change in how individuals felt about the technology's role in the public sector. Many users treated the experience as a routine task without considering the wider effects for governance. This lack of impact suggests that familiarity alone is insufficient to build a stable foundation for digital transformation in state agencies.
Methodology of the Tel Aviv University and King's College Study
Researchers recruited 1,500 workers to participate in a series of simulations that modeled interaction with sophisticated software. These subjects performed tasks where AI provided suggestions, corrected errors, or analyzed complex datasets. Control groups performed similar duties without automated assistance to establish a baseline for comparison. Yotam Margalit noted that the experimental design ensured participants faced the same pressures found in modern office environments. Data collection focused on qualitative assessments of trust alongside quantitative performance metrics during these interactions. Shir Raviv helped structure the feedback mechanisms to capture immediate reactions to algorithmic outputs.
Interaction with the software occurred over several sessions to account for the novelty effect. Initial excitement or skepticism often fades after repeated use, making long-term data points more reliable for academic analysis. Tel Aviv University faculty members reviewed the results to ensure that demographic variables did not skew the primary conclusion. Workers across different age groups and technical backgrounds showed strikingly similar indifference to the software after the initial trial phase. Usage did not breed contempt, but it certainly did not breed confidence.
Impact of Factual Knowledge on Political Decision Making
Information delivery was the second foundation of the experimental framework. While the first group merely used the technology, a second group received detailed explanations of how the AI functioned. These briefings included data on error rates, the logic behind specific calculations, and the human oversight protocols in place. British Journal of Political Science editors highlighted that this educational intervention moved the needle on public opinion. Understanding the "how" and "why" of an algorithm proved far more persuasive than the mere utility of the tool itself. Factual clarity addressed deep anxieties about the "black box" nature of automated systems.
A major new study suggests people's direct experience with artificial intelligence has little impact on their views about its role in government decision-making, while factual information about the technology can sharply shift public opinion.
Public-sector leaders often assume that gradual exposure to technology will naturally lower resistance. Results from this study contradict that assumption. Yotam Margalit and Shir Raviv demonstrated that transparency acts as a catalyst for acceptance. When participants learned about the rigorous testing and specific goals of the AI, their willingness to support its use in government functions increased. Education provides a sense of agency that passive usage cannot replicate. Knowledge transforms the machine from a mysterious interloper into a predictable utility.
The study also separates personal experience from factual knowledge, showing that public trust in AI can change when people understand what a system is actually doing.
Individuals often struggle to reconcile their personal experience with the complex operations of state-level software. Using a chatbot or a basic scheduling tool provides a different psychological baseline than observing a system that determines housing eligibility or tax audits. 1,500 workers in the study showed that the superficial nature of most AI interactions limits their ability to inform political judgment. Personal use is often transactional and narrow. Government use is systemic and carries heavy moral weight. Facts bridge this gap by placing the technology within framework of accountability and law.
Cognitive barriers often prevent people from extrapolating their personal success with a tool to a broader societal benefit. Someone might enjoy an AI-curated music playlist without trusting an AI to manage public health data. Tel Aviv University researchers identified this as a critical disconnect in the current tech discussion. Shir Raviv argued that the public perceives a fundamental difference between convenience and consequence. Factual information serves to reassure citizens that the same standards of justice apply to machines as they do to humans. Trust is a social contract, not a technical byproduct.
Public Trust in AI
The study suggests public AI acceptance depends less on personal novelty than on whether people understand the system's factual basis. That makes transparency a policy tool, not just a communications preference.
For agencies, the finding points toward clearer explanations before deployment rather than public-relations fixes afterward. People appear more willing to accept automation when they can see what information the system uses and where human review remains available.
The practical lesson is that lived experience with a tool does not automatically create trust. Citizens still want to know whether the underlying facts are sound, whether appeals exist and who is accountable when a system gets something wrong.