Public Perception of AI
What the public actually expects, fears, and values in artificial intelligence — and how far that sits from what experts believe.
Public debate about artificial intelligence is loud, and it is mostly conducted by people who build the technology or write about the people who build it. The people who will live with it are surveyed far less often, and usually with instruments too blunt to show structure.
I map that terrain instead of polling it. The method borrows from psychometrics and from risk research: respondents place technologies and concrete applications in a space defined by perceived risk and perceived benefit, by expectation and by value, and the resulting criticality map shows which applications sit in comfortable territory and which are quietly contested. The maps turn out to be stable, interpretable, and considerably more informative than an approval percentage.
Three findings have shaped the rest of the work. Perception is not one-dimensional: an application can be seen as highly beneficial and highly risky at once, and treating that as ambivalence rather than as a tradeoff misreads it. Experts and the public diverge systematically rather than randomly, which means the gap can be characterised instead of merely lamented. And the maps move across cultures in ways that argue against exporting any single country’s AI policy intuitions wholesale.
The instruments are built to be reused. Micro scenarios, criticality maps, and the conjoint platform are all designed so that other groups can run them on their own technologies rather than take my word for the results.
