You are in a group of seven people. You and one other person have each just been given a dollar. The other five got nothing.
Now you are offered a switch. The other person loses their dollar, and the five who got nothing each receive a dollar. Your own dollar is safe either way.
Would you flip the switch?
This is the monetary version of the trolley problem, the famous thought experiment in which you can save five lives by sacrificing one. But here nobody dies, nothing is imagined, and the money is real.
In a study just published in Judgment and Decision Making, my colleagues and I gave this choice to 321 American adults (Vurgun et al., 2026). The project was led by Rozelin Vurgun, then a master’s student in my lab, and now a PhD student at the University of Illinois Chicago.
About 69 percent flipped the switch: They took a dollar away from one stranger so that five others could each have one.
The more interesting question is who made the switch.
Why real money matters
For two decades, the leading account of moral judgment has been Joshua Greene’s dual-process model: Quick, gut-level responses push us toward rules such as “do not harm.” Slower, more deliberate thinking pushes us toward outcomes toward the greater good (Greene et al., 2001). Most of the evidence comes from dilemmas in which people imagine pushing a man off a footbridge to stop a runaway trolley. Nobody has ever faced one. And people who say they would push tend to score higher on measures of psychopathy, so the choice may reflect a weak reaction to harm rather than concern for the five (Bartels & Pizarro, 2011).
Dries Bostyn and colleagues (2019) designed the money version to get around this. In their data, taking from one to give to five was not linked to psychopathy at all.
The 17-month gap
We added something else: time.
Seventeen months before the money task, the same participants had completed a set of tests. Some measured reflection with problems that have an obvious but wrong answer, such as the classic bat-and-ball puzzle (Isler & Yilmaz, 2023). Some measured reasoning ability with pattern puzzles. Others asked people how they like to think: whether they enjoy effort, trust their gut, or change their minds when the evidence says so. They also filled in the Oxford Utilitarianism Scale, a widely used questionnaire with items such as “It is just as wrong to fail to help someone as it is to actively harm them yourself” (Kahane et al., 2018).
By the time the money task came, nobody could have guessed what we were testing.
What predicted the choice
People who had done better on the reflection problems 17 months earlier were more likely to flip the switch. Each extra correct answer, out of five, raised the odds of the greater-good choice by about 19 percent. This was our preregistered test, and it passed; the effect was modest but real.
Two other measures did better. People who scored higher on pattern puzzles and those who described themselves as open to changing their minds were clearly more likely to make the utilitarian choice. When all three were put into one model, these two held up, and the reflection problems dropped out.
So the dual-process model gets support, with a nuance. What mattered was not only the habit of catching a tempting wrong answer, but the ability and the willingness to think things through—and people who flipped the switch took longer to decide, not less. This was not impulsiveness.
What did not predict the choice
The Oxford Utilitarianism Scale did not. Neither of its two parts—concern for everyone’s welfare and acceptance of harm for the greater good—was related to what people actually did.
Worse, scores on the scale rose with a preference for intuitive thinking and fell with a preference for reflection. The welfare items also correlated with almost every moral justification we asked about, including rule-based ones. That is close to the opposite of what a measure of utilitarianism should show.
The scale is not useless. But saying you care about everyone is not the same as taking a dollar from someone to prove it. Self-reports of moral values need to be checked against behavior, and here they failed the check.
How faith changes the picture
The result I find most interesting was not preregistered, so treat it as a lead rather than a conclusion.
Half of our sample were Christian believers, and the other half were atheists or agnostics. Religious people were somewhat less likely to flip the switch, in line with earlier findings that believers favor rule-based choices in moral dilemmas (Saroglou & Craninx, 2021).
But the link between reflection and the greater-good choice depended on belief. Among non-believers, more reflective people were much more likely to flip. Among believers, reflection made little or no difference.
This fits something I have argued here before: Reflection is not a truth machine. It is a tool that operates according to whatever standards of good reasoning a person has absorbed (Yilmaz & Isler, 2025). For a non-believer, that often means weighing evidence and outcomes. For a believer who sees divine command as the source of moral truth, careful thought can just as easily strengthen the rule against taking what belongs to someone else.
As we put it in the paper, reflection “can be recruited to bolster an already favored conclusion rather than to evaluate it.”
Takeaways
This is one task, with one dollar at stake, in one country. Our data are correlational, and the religion result needs a direct test.
Still, three things seem clear:
- Thinking ability and open-mindedness predict a moral choice more than a year later.
- Questionnaires about moral values can fail to predict what people do.
- Whether careful thinking pushes someone toward the greater good may depend on what they believe careful thinking is for.
So the next time you hear that smarter or more reflective people are more “utilitarian,” ask two questions: Utilitarian in what they say, or in what they do? And reflective by whose standards?
