When Science Isn’t Gaslighting You After All
Science can save lives by overturning our deepest, most well-meaning assumptions about the causes of harm
Scientists are sometimes blamed for delaying accountability and public knowledge on clear harms and injustices. In some cases the blame is fair — when a powerful actor demands impossible scientific evidence on a harm that everyone can see.
But if science is a source of delay, why rely on it at all?
I’m a computer scientist and behavioral scientist at Cornell University who works together with the public to study the impact of technology on society, and work together to test effective interventions for a safer, fairer digital environment. I’ve worked alongside the public to investigate issues including online violence, algorithmic discrimination, and youth mental health.
In every project, community partners and I have a conversation about the norm in science — where we commit to following the evidence wherever it leads. I am always grateful and inspired when people work with us when they are being ignored by powerful institutions and are hoping that science will validate their claims. A successful partnership also requires us all to prepare to learn something we didn’t expect.
I also have these conversations with students at Cornell, in the class I teach on investigating AI failures. This week, I introduced students to one of the most horrific cases of unscientific evidence run amok: research on the causes of sexual assault.
In the US according to the CDC, nearly half of women and more than 1 in 6 men have experienced some form of contact sexual violence in their lifetimes, with more than 1 in 5 women experiencing completed or attempted rape. Given this endemic, society-wide harm, many schools and colleges put significant resources into understanding sexual assault and trying to prevent it.
Unfortunately, many interventions on sexual assault are based on a single study with a small, non-probabilistic sample that promoted the “serial rapist” model. The flawed paper claims that a small minority of atypical people are responsible for over 90% of sexual assault on campuses- that it is the result of “a few bad apples.” If you have ever taken a bystander training, you have been trained in the serial rapist model and have probably heard this number.
The problem is that the “bad apples” theory is not accurate. Multiple large-sample longitudinal studies have not found evidence for the serial rapist hypothesis. While some people are repeat offenders, they find evidence that sexual assault is more endemic to society and perpetrated by a wider range of people. Yet findings from that single, flawed 20+ year old study refuse to go away. As a result, many people and organizations have held onto a fundamentally mistaken framework, spending countless millions of dollars and leaving people unprotected for decades.
How can a debacle like this happen? Psychologists Ana Gantman and Elizabeth Levy Paluck recently published a paper finding that the serial rapist model, even though it’s not true, is associated with people’s belief in a just world and related beliefs. Confronting the truth would be too painful, forcing many people to rethink assumptions about our communities. So people latch on to bad science and throw millions of dollars at misplaced efforts that provide cover for the ongoing scourge of sexual assault.
Confronting the truth would be too painful, forcing many people to rethink assumptions about our communities. So people latch on to bad science.
One of the things I admire the most about Gantman and Paluck’s work is that they didn’t stop after pointing out the problem — they have been tireless at working with communities to test interventions to change the climates that enable sexual assault. By shedding light on misconceptions, they have created new space for solutions.
On its best days, science helps us ground our beliefs in what is observably happening in the world. Whenever someone pressures me to publish data that confirms my assumptions, especially when I feel a sense of urgency and moral outrage, I think about the researcher whose quick, unreliable study on sexual assault set back public health by decades.
Yes, it’s bad when slow science delays a search for solutions. But it’s sometimes even worse when fast, unreliable science creates a false trail for well-meaning people by misleading the public about harms that need urgent solutions.
It can feel like a miracle when scientists and hurting communities find common cause in a search for the truth, and I’m filled with wonder and hope every time it happens.
Further reading:
- COMM 4940: Investigating AI Failures
- Gantman, A. P., & Paluck, E. L. (2025). What is the psychological appeal of the serial rapist model? Worldviews predicting endorsement. Behavioural Public Policy, 9(2), 461–476.
- Pierson, E., Li, S. (2015) A better way to gauge how common sexual assault is on college campuses. The Washington Post
- Rosenberg, M., Townes, A., Taylor, S., Luetke, M., & Herbenick, D. (2019). Quantifying the magnitude and potential influence of missing data in campus sexual assault surveys: A systematic review of surveys, 2010–2016. Journal of American college health, 67(1), 42–50.
- Porat, R., Gantman, A., Green, S. A., Pezzuto, J. H., & Paluck, E. L. (2024). Preventing sexual violence: A behavioral problem without a behaviorally informed solution. Psychological Science in the Public Interest, 25(1), 4–29.
- The Monument Quilt — oral histories from the public healing space by and for survivors of rape and abuse established on the National Mall in Washington D.C. in the summer of 2019, and recorded by StoryCorps.
- Matias, J. N., & Price, M. (2025). How public involvement can improve the science of AI. Proceedings of the National Academy of Sciences, 122(48), e2421111122.
- Orben, A., & Matias, J. N. (2025). Fixing the science of digital technology harms. Science, 388(6743), 152–155.
