Method, Mind, and Morality: How People Make Sense of Artificial Intelligence
This paper analyzes the sensemaking dynamics of AI through a mixed-methods study of media and interviews with professionals, proposing a framework of three contested debates—method, mind, and morality—that shape how society interprets and responds to the rapid advancement of artificial intelligence.
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
Artificial intelligence has moved from the pages of science fiction into the fabric of daily life, appearing in everything from the news feeds we scroll through to the tools we use at work. Yet, as these systems become more capable, a profound question remains: how do human beings actually understand them? We do not just process data; we process meaning. To make sense of a new technology, people rely on mental shortcuts and shared stories, known in social science as "frames." These frames act like lenses, organizing complex information into a shape our minds can grasp. When a technology changes rapidly, as AI is doing now, these lenses can clash, creating confusion about what the technology is, how it was built, and what it should be allowed to do. Understanding these mental models is not just an academic exercise; it is essential for navigating a future where machines increasingly shape our economy, our laws, and our relationships.
To uncover how people are making sense of this shift, researchers Jacy Reese Anthis, Erik Brynjolfsson, and James Evans conducted a massive investigation that combined the broad sweep of digital data with the intimate detail of personal conversation. They did not start with a hypothesis to prove, but rather let the data guide them. First, they analyzed millions of documents, including hundreds of thousands of newspaper articles and over a million posts from social media, spanning from 2018 to 2024. They used computer tools to identify the recurring themes and words that people were using to talk about AI. Then, to understand the human reasoning behind those words, they conducted 57 in-depth interviews with AI professionals, including managers, researchers, and engineers, in 2021 and again in 2023. This timing allowed them to witness how thinking evolved before and after the sudden surge of powerful new AI systems that captured global attention.
The study revealed that people are struggling with four specific cognitive challenges as they try to keep up with AI. The first is simply the speed of change; the technology is advancing so fast that even experts feel they cannot stay fully informed, with one manager noting that their skills can become outdated overnight. The second challenge is communication; professionals find it difficult to explain their work to family, friends, or colleagues who do not share their technical background, often forcing them to simplify complex ideas into terms like "selling data" or "software engineering" just to be understood. The third challenge involves responsibility; when AI causes harm, such as spreading misinformation or making biased decisions, it is incredibly difficult to pinpoint who is to blame, as the technology is built by many people using data from many sources. The fourth challenge is making trade-offs; people struggle to balance competing goals, such as the desire for efficiency against the need for fairness, or the push for innovation against the need for safety.
To navigate these challenges, the researchers found that people are locked in three major debates, or framing contests, that define how society views AI. The first debate concerns the method of creation: is AI built like a top-down expert system, where humans code strict rules, or is it a bottom-up system that learns patterns from massive amounts of data on its own? The second debate concerns the mind of the machine: is AI merely a passive tool, like a calculator, or is it an active agent, a "digital coworker" with its own form of intelligence? The third debate concerns morality: should we slow down the development of AI to manage the risks, or speed it up to capture the benefits? These are not just technical arguments; they are stories people tell themselves to decide how to act.
The researchers discovered that the way people frame these debates has real-world consequences. When experts try to communicate nuanced ideas, the message often gets simplified as it spreads. For instance, a warning about the dangers of advanced AI intended to slow down development might be heard by investors as a signal that the technology is powerful and valuable, prompting them to accelerate development instead. The study suggests that the "productivity paradox"—the observation that new technologies often take years to show economic benefits—might be partly caused by these social and cognitive struggles. Just as factory managers once needed time to redesign their floors to use electric motors effectively, society may need time to agree on the right mental models for AI before it can be fully integrated.
Ultimately, the paper suggests that the future of AI will be shaped not just by code and hardware, but by the stories we tell about it. The researchers found that while some frames, like the idea of AI as a human-like companion, are deeply embedded in our culture, others, like the specific nature of AI's intelligence, are still up for grabs. This uncertainty creates an opportunity. By understanding the mental models people are already using, designers and policymakers can better guide the conversation, helping society to adopt frames that align with safety and public benefit. The study does not claim to have solved the problem of AI safety, but it provides a clear map of the cognitive terrain we must cross to get there, showing that how we think about machines is just as critical as how we build them.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.