Based on Smaldino & McElreath (2016), “The Natural Selection of Bad Science”

Introduction: Fraud Doesn’t Begin With Villains—It Begins With Incentives

When people imagine scientific fraud, they picture caricatures:
a rogue scientist faking data to become famous; a malicious figure cooking numbers for personal gain.

Reality is much more subtle—and much more disturbing.

Fraud evolves.

Just like biological traits arise in response to environmental pressures, fraudulent behaviors arise within scientific ecosystems because those behaviors confer competitive advantage under certain institutional conditions. Smaldino & McElreath’s paper argues that the same incentives that select for poor methodological rigor also select for increasingly bold forms of cheating.

This evolutionary view challenges the idea that individual misconduct is the root of the crisis. Instead:

Fraud is an adaptive response to misaligned incentives, not a personal flaw in an otherwise healthy system.

In this post, we explore:

How small methodological shortcuts evolve into systemic fraud
Why fraud emerges even among people with good intentions
Historical and modern examples of fraud evolution
How fraudulent strategies spread within academic lineages
Why policing fraud is so difficult in an ecosystem selecting for it
And critically: how we can redesign incentives to prevent fraud’s natural selection

1. The Continuum of Cheating: From Innocent Flexibility to Full Fabrication

Fraud rarely begins with outright forgery. It evolves gradually through a dangerous continuum:

Stage 1: Innocent flexibility

Researchers try multiple statistical models because “they want to understand the data better.”

Stage 2: Selective reporting

Negative results are dropped “because the journal won’t accept them.”

Stage 3: HARKing (Hypothesizing After Results are Known)

Researchers rewrite hypotheses post-hoc to match significant results.

Stage 4: Data massaging

Removal of outliers that “don’t make sense,”
or reclassifying categories to achieve significance.

Stage 5: Fabrication-lite

Inventing a few missing values, adjusting means slightly, or copying data points to “fix noise.”

Stage 6: Full data fabrication

Creating entire datasets from imagination.

The key insight is this:
At each step, competitive advantage increases while detection risk remains low.
Evolution always favors strategies with the highest payoff relative to cost.

Misaligned incentives—publish or perish, novelty over accuracy, prestige over honesty—act as selective pressures moving individuals along this continuum.

2. Why Good People Drift Toward Bad Science

People do not enter science as cheaters. They enter as idealists.
But as evolutionary biologists know, behavior adapts to the environment.

Three forces push researchers toward unethical behavior:

2.1 Selection for productivity over accuracy

A researcher who produces 12 papers a year—thanks to flexible methods or data manipulation—is more likely to get:

job offers
grants
tenure
speaking invitations
media attention

Meanwhile, the careful, slower researcher is deemed “less productive.”

This is pure Darwinian selection, not moral selection.

2.2 Lack of punishment mechanisms

In nature, cheating thrives when policing is absent.
In academia:

Fraud detection rates are extremely low
Replication rarely occurs
Retractions are rare and slow
Institutions protect successful researchers
Journals avoid scandals to protect their reputation

Low policing + high reward = the perfect conditions for cheating to thrive.

2.3 Cognitive dissonance and rationalization

Once minor cheating yields rewards, researchers begin to rationalize:

“Everyone does it.”
“The result is basically true.”
“This helps me survive until tenure.”
“The reviewers won’t understand anyway.”
“I know the effect is real—I just need cleaner numbers.”

This psychological lubricant allows unethical behavior to seem justified.

3. Fraud Evolves Because It Works

Smaldino & McElreath’s core argument is simple and devastating:

The system selects for those who succeed—not those who are honest.

Every generation of researchers learns from the successful.

And who is successful?

The ones who:

publish frequently
produce flashy claims
get into prestigious journals
secure big grants
attract media coverage

If these successes are achieved through questionable practices, then those practices become heritable—not genetically, but culturally, through lab training and mentoring.

4. Fraud Spreads Through Academic Lineages

Just as biological traits spread through reproduction, research practices spread through academic genealogy.

4.1 “Descendants” adopt their mentors’ strategies

If a PI produces statistically improbable results regularly, their trainees absorb:

their data methods
their publication strategies
their analysis shortcuts
their attitude toward p-values and significance
their tolerance for exaggeration

This creates academic “lineages” with distinct methodological cultures.

Evidence: David L. Stern & the “lineage effect”

Stern’s work showed that even fruit flies inherit certain behavior patterns culturally across generations.
Labs do too.

Good practices and bad practices propagate through lineages.

4.2 Fraud clusters geographically and institutionally

Just like infections spreading in populations, fraud patterns cluster:

similar manipulation techniques
same statistical artifacts
same impossible distributions
same writing styles
same figure duplications

These clusters reveal that fraud is not random—it is learned.

5. Historical Examples: Fraud Evolution in Action

Fraud is not new, but it is increasingly detected in clusters, consistent with evolutionary models.

5.1 The Cyril Burt IQ scandal

Cyril Burt published “twin studies” claiming extremely high heritability of intelligence.
Later, investigators found:

nonexistent coauthors
fabricated correlations
copied data patterns

For decades, he thrived.
His fraudulent work shaped education policies.
And his students inherited his methods.

This is classic evolutionary propagation: successful phenotype → expanded lineage.

5.2 Anil Potti (Duke University cancer genomics)

Potti published numerous high-impact cancer biomarker papers.
Later:

analyses showed fabricated patient data
bioinformatic methods were manipulated
clinical trials were influenced

His lab’s success created a generation of scientists trained on toxic practices.

5.3 Diederik Stapel (Social psychology)

Stapel produced extremely clean datasets that were “too perfect.”
His fraud persisted because:

he trained students with similar data expectations
his results matched reviewers’ theoretical biases
replication was rare

The ecosystem protected him.

5.4 Yoshinori Watanabe (Cell biology)

Watanabe’s lab was caught manipulating blots and fluorescence images.
Investigations revealed:

systemic training in visual data manipulation
multiple students involved
institutional reluctance to punish

Fraud had become a lab culture, not individual misconduct.

6. Why Fraud Thrives in the Current Scientific Ecosystem

Fraud spreads because the ecosystem selects for it.
Smaldino & McElreath highlight several systemic pressures:

6.1 Lack of replication removes constraints

Replication is the natural predator of fraud.
But when replication is rare, fraud proliferates.

6.2 High competition intensifies selective pressure

When survival depends on out-producing rivals:

statistical flexibility becomes adaptive
selective reporting becomes strategic
fabrication becomes tempting

This is akin to bacteria evolving antibiotic resistance under selective pressure.

6.3 Journals reward “too good to be true” results

Fraudsters know what reviewers want:

large effect sizes
perfect curves
clean p-values
dramatic conclusions

This mirrors sexual selection in nature: whatever trait is preferred, individuals evolve to exaggerate it.

6.4 Institutions protect high-performers

Universities benefit from:

prestige
funding
high-impact publications
media attention

They often resist investigating misconduct because the fraudster benefits the institution.

This is group-level incentive misalignment.

7. The Fragility of Policing Mechanisms

Unlike biological evolution, which often has built-in constraints, scientific culture has weak policing:

7.1 Journal peer review rarely checks raw data

Reviewers lack time, expertise, or access.

7.2 Retractions take years

Retraction Watch tracks retractions that took decades.

7.3 Whistleblowers face retaliation

Whistleblowing can destroy careers.

7.4 Detection methods lag behind fabrication techniques

For instance:

easy digital manipulation of images
generative AI for synthetic data
deep statistical obfuscation
complex bioinformatic pipelines

The result:
Fraud evolves faster than policing.

8. Can Fraud Ever Be Eliminated? Evolutionary Theory Says No—But It Can Be Minimized

In natural ecosystems, cheating strategies never disappear entirely.
But they can be controlled by making cheating:

less rewarding
more risky
more detectable

The same must be done in science.

8.1 Increase the cost of cheating

mandatory raw data and code availability
unblinded access to analysis pipelines
random replication audits
statistical anomaly detectors
funding agency spot checks

8.2 Reduce the rewards

prioritize quality over quantity
reward transparency
value incremental progress
shift journal prestige toward replicable work

8.3 Enhance policing

fast-track retractions
strong whistleblower protections
specialized forensic-statistics units
replication consortia to investigate suspicious papers

8.4 Change cultural expectations

The real transformation begins when lab culture shifts from “show me significance” to “show me validity.”

Conclusion: Fraud Is Not a Disease of Individuals—It Is an Evolutionary Outcome of the System

This is the most sobering conclusion of Smaldino & McElreath’s work:

Fraud is inevitable in a system that rewards fraudulent strategies.

Unless incentives change, the evolution of scientific misconduct will continue—and accelerate.

Fraud is not merely a failure of morality.
It is a failure of ecology.
A failure of institutional design.
A failure of evolutionary pressures.

We can restore integrity only by reshaping the selective landscape so that:

honesty becomes adaptive
replication becomes central
transparency becomes mandatory
quality becomes rewarded

Only then will the evolution of fraud slow—and perhaps stabilize at manageable levels.

Friday, February 6, 2026

Post 8: The Evolution of Fraud—Why Misaligned Incentives Make Cheating Inevitable