Researchers in the United States have unveiled a new framework for understanding how people decide on political choices, revealing that the majority rely on simple mental shortcuts rather than weighing every available feature. The Berkeley-based study shows that the decision rule people use can shift depending on the topic and the complexity of the choice environment, challenging assumptions that survey analyses capture all deliberation.
In political science, choice-based conjoint experiments are widely used to gauge public opinion on complex issues. Participants are shown side-by-side profiles of options—such as candidates or policy proposals—with varying features, and asked to pick a preference. From these choices, researchers infer how much each feature sways voting behaviour across the population.
The researchers describe four distinct decision strategies that differ in cognitive demand. The most thorough is Weighted Additive, a rational process in which a person weighs every attribute and sums them to pick the top-scoring option. A simpler shortcut, Tally, counts the number of preferred traits in each option and selects the one with the higher tally. The Take The Best rule is the simplest, directing a choice based on the single most important feature, only consulting a second feature if the top trait ties the options. The Contaminant category captures random guessing, equivalent to a coin flip.
“We were interested in the question of how people go about making political choices,” Banavar told PsyPost. “When we estimate complex constructs like political attitudes, a lot of analysis methods implicitly assume that people weigh everything in front of them. We wanted to test whether that was true, and if not, to find how they were deciding instead.”
Banavar, a postdoctoral scholar in the Travers Department of Political Science at the University of California, Berkeley, explained that much political attitude research is conducted at the aggregate level, which “collapses a lot of the variation we see in people: both in terms of their actual preferences and in terms of how they make choices based on those preferences.” The study set out to uncollapse the aggregate by building a model and designing experiments to test how individuals actually decide.
A central finding is the prevalence of heuristic use. Banavar said: “We were interested in the question of how people go about making political choices.” She added that “the big takeaway is the prevalence of heuristic use and how the majority of individuals use informational shortcuts… people are not exhaustively weighting and evaluating all the information presented to them.” She also stressed a nuanced result: “How someone processes information is not a trait: just because I may choose a candidate based on one factor doesn’t mean that I’ll choose a climate policy based on one factor. This means we shouldn’t put people into buckets based on how they make political choices.”
The team’s model was tested in two experiments involving 3,762 American adults recruited via Prolific, with quotas matching national demographics for age, sex and political affiliation. Participants completed 90 choices divided into three blocks, covering three common political topics: two hypothetical immigrants seeking U.S. entry, two Senate primary candidates, and two national climate policy packages. Each option varied across six attributes with two possible values.
In the first experiment, 1,964 participants faced profile sets that either never contained identical traits or included one to three matching features. This design aimed to see whether forcing respondents to look beyond their top attribute would promote more thorough evaluation strategies.
Across both conditions, the Take The Best rule emerged as the most common strategy in all three political domains. In the no-ties condition, about 54% of participants were classified as Take The Best users, roughly 30% weighed all features in a Weighted Additive fashion, and about 14% followed the Tally approach. When ties were possible, Take The Best use fell to around 45%, while rational Weighted Additive use rose to about 40%.
The researchers found that many people did not stick to a single approach. Between 67% and 76% of participants were classified as using different strategies for different issues, with around 10% employing a completely different strategy across all three domains. Topic order was randomised, and there was no evidence of fatigue or increased guessing over the course of the rounds.
“The big takeaway is the prevalence of heuristic use and how the majority of individuals use informational shortcuts… This is an important finding – people are not exhaustively weighting and evaluating all the information presented to them,” Banavar observed. “However, I think a bigger takeaway is that people tend to use different strategies depending on the political domain.”
In the second experiment, 1,798 participants were asked to rate each profile on a 1-to-7 scale in addition to making a comparative choice. For those given ratings, fewer took the Take The Best approach across the three domains, though it remained the most common strategy. The ratings condition also produced more individuals classified as employing more thorough strategies, alongside a small uptick in random guessers, a shift the authors cautioned may reflect waning engagement in some respondents.
Response times provided a supplementary check on the classifications. Those labelled as using Weighted Additive took the longest to respond, followed by Tally users, with Take The Best respondents answering fastest. When ties were present, only the Take The Best group showed slower responses; other groups’ times remained essentially unchanged. The authors interpreted this as evidence that the classifications reflect real differences in mental processing, though they stopped short of claiming to observe thought directly.
The researchers acknowledge that the findings have limits. The experimental profiles used two-valued features, which are simplifications of real-world political choices that often involve more layers and continuous values. They emphasised that the results do not imply laziness, noting that finite cognitive resources can be efficiently deployed to maximise payoff.
Demographic variables offered little predictive power about strategy choice. Across race, gender, education and political affiliation, associations were generally weak and often not statistically distinct after correction for multiple tests. Some modest patterns emerged—older participants showed a slight tendency toward Take The Best, and those who perceived a topic as less complex tended to favour shortcuts—but these effects did not consistently hold.
The authors also warned about the limitations of their classification approach. A person who weights every feature but places overwhelming emphasis on one could be misclassified as a Take The Best user, and simulations suggested that random guessers could be mistaken for Tally users. Banavar stressed that the model is a classification tool, not a direct readout of cognitive processes.
Looking ahead, the researchers propose extending the framework to features with more than two values, exploring why people choose particular strategies, and modelling response times alongside choices. The study, “Classifying strategy use in multiattribute subjective choice: Application to conjoint experiments in political science,” was authored by Nidhi V. Banavar and Kirk Bansak.
