IAH 206 Part I Review

Define Deductive Reasoning

Form is valid; truth preservation guaranteed; premises true = conclusion will be true

Ampliative

Adding new information beyond what is already contained in a concept (generate new knowledge)

Deductive Reasoning (Types)

Modus ponens and modus tollens

Deductive Reasoning (Role)

Can be used to draw out implications of theories/new predictions to test theories

Modus Ponens

Deductive Reasoning; if p, then q, p exists, therefore, q

Modus Tollens

Deductive Reasoning; if p, then q, not q, therefore, not p

Modus Ponens (Example)

(1) If it rains, then the ground is wet. (2) It is raining. Conclusion: Therefore, the ground is wet

Modus Tollens (Example)

(1) If it is raining, then the ground is wet. (2) The ground isn’t wet. Conclusion: Therefore, it is not raining

Deductive Inference (Scientific Reasoning)

Used to find what follows from a theory/hypothesis; good for developing predictive tests

Define Inductive Reasoning

Truth preservation is NOT guaranteed; even if the premises are true, conclusion may not be; deductively invalid

Inductive Reasoning (Concept Example)

(1) If p, then q (2) q, q, q, q, q, q,… conclusion: therefore, q?

Inductive Inference (Scientific Reasoning)

An ampliative inference form; how evidence supports a theory/hypothesis; no guaranteed truth

Inductive Reasoning (Role)

Can be used to generate broader claims or to show evidential support for theories

Inductive Reasoning (Real-life example)

Buying a gift for your mother: You remember she likes dark chocolate, but she also doesn’t like dark chocolate with coconut flakes, you know a nearby candy store that sells dark chocolate with many fillings, you buy an assortment of dark chocolate candies with different fillings, and you notice she seems to prefer candy filled with caramel, so you keep that in mind next time you want to get her a present you know they’ll appreciate.

Why do we need scientific theories/hypothesis?

To direct out efforts at testing, to know which data is relevant and thus which data to collect, and to observe

Where does scientific theories/hypothesis come from?

Creative insight, happenstance (luck), cultural background

Semmelweis (Hypotheses)

Miasma/atmospheric conditions, overcrowding, medical student rough handling, psychological dread generated by the priest’s bell, birth position, Kolletschka’s death/cadaveric matter.

Uncertainty and Reliability in Science

Evidence supports theories through inductive inference forms. Theories can never be 100% truely known. New evidence and/or better theories can challenge current theories.

Why is science the most reliable form of empirical knowledge?

The community is continually testing and critiquing its work. Key source of science’s objectivity and reliability.

Problem with Evidential Suffiency?

Not good for short-term. When is the evidence enough? Scientists’ disagreements with each other on these grounds

Semmelweis (Aftermath)

Insisted on a single cause for childbed fever. Colleagues rejected his view and demands for better handwashing. Restrained in an insane asylum and died from sepsis (childbed fever). Never got the theory right. Pasteur and Lister developed germ theory of disease and better disinfectant methods.

Merton’s Social Structure of Science

Ethos of science as institutional norms, not necessarily individual norms (cultural structures of science reflect these and shape individual behaviour; individuals can violate these norms)

Is science independent of society?

Just produces truth which drops into society

Is science full embedded within society?

Just reflects the norms and ideas of society

Is science distinctive in its practises?

Within society, but also has its own culture.

Merton’s Ethos of Science (CUDOS)

Universalism, Communalism/Communism. Disinterestedness, Organised Skepticism. These are distinctive norms of science, that may or may not fit with the broader society.

Universalism

Shouldn’t matter who does the science for assessment of the science.

Basic: No exclusionary barriers

Communalism/Communism

No private ownership in science; just recognition and esteem (priority disputes)

Basic: sharing results (and credit) is essential

Disinterestedness

NOT about individual’s lac of interest in science or results; institutional structures demand critical scrutiny of results; bases for accepting a claim are not found in the interests of the scientists pursuing or supporting the claim; public accountability to peers

Basic: culture of critique is needed, critique based on evidence, theory, method

Organised Skepticism

Ongoing scrutiny; nothing held above scrutiny (no dogmas)

Basic: no dogmas; nothing above critique

Interrelationships among the norms (U/DOS)

Universalism ensures broad base for criticism needed for disinterestedness and organised skepticism

Interrelationships among the norms (C/DOS)

Communalism ensures sharing of results needed for disinterested and organised skepticism

Interrelationships among the norms (OS/U)

Organised skepticism supports universalism

Interrelationships among the norms (D/OS/U)

Disinterestedness provides a way for organised skepticism to be maintained and reinforces universalism

Does universalism fit with democracy?

Yes. Both require equality of opportunity and impersonal judgment.

No. Democracies often prioritise local, national, or political identities over universal ones

Does communalism fit with democracy?

Yes. Both rely on transparency and the open sharing of information

No, capitalist democracies actively incentivize innovation through patents, intellectual property laws, and corporate secrecy.

Does disinterestedness fit with democracy?

Yes. Both require accountability and service to the public interest over personal gain.

No. Democracies run on political lobbying, special interests, and funding agendas

Does organised skepticism fit with democracy?

Yes. Both depend on free speech, debate, and questioning authority.

No. Democratic politics often demand quick consensus and emotional appeal over prolonged doubt

Differences between norms of science and norms of democratic politics

Democratic politics and demands for loyalty

Scientific expertise can feel exclusionary to public

Voting in politics

Cultures of criticism and response

Similarities in norms of science and norms of democratic politics

No ones owns the idea in politics or science

Broadest base for debate produces the best results in the long run

Open debate and critique is an essential practise

Milikan’s Oil Drop Experiment

Purpose: Assessing the charge of the electron and assessing accuracy of the Stokes’ Law correction.

Didn’t carry out calculations on drops too small (Brownian motion) or too large (would fall too quickly to measure accurately).

Preferred drops that changed direction

Threw out data

Milikan and the Social Functioning of Science

Won a Nobel Prize. Became head of Cal Tech. Kept cedit from his grad student. Wrote against hiring women. Wrote racist terms about Jewish scientists and supporter of eugenics (no more than one Jewish scientist was hired by Cal Tech per year). Cal Tech removed his name from the buildings, but not for scientific fraud.

What is scientific fraud?

Fabrication, falsification, plagiarism

The Problem with Fraud

Falsifies record of scientific data (which scientists presume to be accurate). (Kicking out bad runs doesn’t commit fraud, as long as it is clear why it is a bad run).

Define Fabrication

Making up data

Falsification

Changing data to suit preferences (includes image manipulation)

Plagiarism

Claiming someone’s words as your own

Scientific Fabrication/Falsification (Intent)

Always deliberate; intending to deceive what happened in the lab; faking the records

Conditions for Fabrication/Falsification Fraud

Career pressure; “knowing” how the study should work out; complications with reproducibility for the study

Plagiarism (intent)

Can be unintentional

How is Fraud Detected?

Failed replications, lab colleagues (including grad students) who see something amiss, exceptional productivity arouses suspicion, some people look for it, peer review(??)

Role of Peer Review

Good at: identifying sloppy inferences and inapt methodologies and weeding out poorly conceived projects.

Not good at: detecting fraud

Why not?: Peer reviewers do not reproduce experiments themselves. They read reports from other scientists’ work. Those reports need to be accurate. Fraudulent reports aren’t.

Questionable Research Practises (QRPs)

P-Hacking, hypothesizing after the results are known (HARKing), ignoring outlier data, selective reporting of data

P-Hacking (QRP)

Scientist keeps testing for statistically significant correlations until some arise (if p < 0.05 for statistical significance, it’ll happen by chance 5% of the time).

If correlation is promising, another experimental test is needed.

Registered reports alleviate the pressure to do this.

Ignoring Outlier Data (QRP)

Sometimes for good reason (Milikan). Sometimes it’s the wrong thing to do (discovery of ozone hole).

Selective Reporting of Data (QRP)

Pick subject populations where clinical trials worked (cherry-picking).

Trial registries were created to fight this problem

Registered Reports to the Rescue?

Proposed method only goes out to peer review (they make suggestions for improvement). Journal guarantees publication for accepted methodological proposals (no need for significant results).

Define Registered Reports

Distinct publication model

Advantages of Registered Reports

More failures get reported

Removes pressure to p-hack to get publications

Peer review occurs before money is spent

More honest depiction of what happens in science

Piltdown Man

1912 Charles Dawson claims to have found skull of missing link between ape and humans. Other scientists not convinced. Radioisotope dating proved it was fraud in 1953. Found to be a regular fraudster.

What links these scientific fraud practises?

Taking credit for work you didn’t do

Whether you think what you wrote is true is irrelevant

Bright lines delineate when something is fraud

Conditions for Scientific Fraud: Career Pressure

Publish or perish

Research production and discovery

Conditions for Scientific Fraud: Thinking the data would have turned out this way

Scientists not trying to deceive people in this way.

Rather deception is about whether the work actually was done or came out this way

Conditions for Scientific Fraud: Difficulty in Replication

Gives cover to possibility that work would have not turned out this way

Three Issues With Fraud

Intention: for fabrication and falsification, intention matters

Collaborator responsibility

Lab head responsibility

Intention (issue with fraud)

Was the image publication an error or deliberate?

Often determined by patterns over time

Papers are retracted regardless

Why is intention necessary for this kind of fraud?

Collaborator Responsibility (issue with fraud)

Lots of authors on scientific papers

Who is responsible for the error or fraud

Lab Head Responsibility

Often the lab head is one of the authors (even if they didn’t really help with the work)

Should they be held partially responsible for retracted work?

Fraud in Biomedicine at Cal Tech (Kumar)

Kumar: Faking Figures

• Random noise shouldn’t be repeated so clearly

• Defense: Didn’t know that this was wrong to do

• Defense rejected because Kumar was not a beginning student

Fraud in Biomedicine at Cal Tech (Urban)

Urban: Phony data in submitted paper

• Planned to swap out phony data for real data once paper was accepted

• “Knew how the experiment would come out if he did it properly”

• Problem: peer reviewers needed to see real data

• Time pressure led him to do it

Fraud in Physics: Schön

Organic Semiconductors

• Got remarkable results

• Published a lot

• Too remarkable

• No one could get any of these to work

• No data or samples stored

Fraud in Physics: Ninov

Element 118

• Found experimental results that confirmed theoretical predictions

• Others (with more powerful equipment) didn’t

• Theory turned out to be wrong

• Ninov was the only one who could have altered all the data to make it fit the predictions

Prevalence of Scientific Fraud

Fraud is rare in science

Rates of retraction are rising

Novel Forms of Fraud

Fraudulent peer review, fraudulent human subject approval (fake the IRB approval papers), fraudulent special issue editing (help your friends get a publication line), and fraudulent citation practises (citation trading).

Structures to Combat Fraud

Research Integrity Offices, better research ethics training, change culture of science, stronger punishments, and more attention to fraud

Research Integrity Offices (Combat Fraud)

Where you would report fraud

Better Research Ethics Training (Combat Fraud)

In-person training more impactful

Change Culture of Science (Combat Fraud)

Registered reports

Less grant pressure

Stronger Punishments (Combat Fraud)

Higher fines

Jail time

Permanent job loss

More Attention to Fraud (Combat Fraud)

Retraction Watch

Better journalistic coverage

Marc Hauser Fraud

Falsified data on primate behaviour.

Trying to show that some primates had similar cognitive abilities of great apes

Role of Grad Students/Research Assistants in Hauser Fraud

• Noticed that coding animal bheaviour didn’t match what they saw in videos

• Hauser refused to address discrepancies or bring in additional eyes

• Grad student whistleblowers brought concerns to Harvard admin

Lab Insiders in Hauser Fraud

Crucial.

• Replication challenging

• Maybe Hauser’s primates were smarter? More relaxed? Better motivated?

Review of Data, Tapes, and Papers in Hauser Fraud

• Took from 2007 to 2012 for investigation to be completed

• Hauser resigned in 2011 from Harvard

Wakefield, the MMR Vaccine, and Autism

1998 Lancet study

• 12 children

• claimed link between the MMR vaccine, colitis, and autism

• claimed to find clear temporal association

Key to ramping up concerns about vaccines and autism

• lots of press coverage

• Lancet, a well-respected medical journal

• if data was true = important finding and should have directed more research

• vaccination rates began to fall (UK), first death from measles since 1992 occurs in 2006

Never replicated

Wakefield asked to leave in 2001 for failing to pursue replication

Takes another 10 years to lose medical license and have full paper retraction

Wakefield Fraud

Brian Deer (journalist) exposes the fraud

Discovers falsification of each child’s medical conditions, or timing of vaccine, or both

Falsification crucial to making it appear there was a relationship between MMR, colitis, and autism

Why is Image Fraud Important?

Key source of data in some scientific fields

Can be crucial to scientific claims

AI and deep fake image fraud

Benford’s Law of Digit Distribution

Need a large sample of numbers

Need distribution over several orders of magnitude

Can check to see if counts line up

The Rise of the Fraud Sleuth

Center for Scientific Integrity (Retraction Watch, medical Evidence Project (2025))

Some scientists now focusing on fraud detection and retraction efforts (James Heathers, Ivan Oransky, Elisabeth Bik)

How to Reduce Fraud Rates?

• Reduce simplistic academic metrics (quality over quantity; problematic: publication counts, citation counts, H-factor)

• Change publication practises to registered reports (peer review for method only, publication assured regardless of results)

• Pursue replication projects

• Staff fraud detection at journals

• Increase whistleblower protections and rewards

• Strengthen punishments

Define Self-Deception

NOT fraud

Scientists believes they are doing science properly

Scientists believes in their results and that they accurately reflect what has happened in the study

Blondot and N-Rays

1903, he announces discovery of N-rays. 1904, he receives Prix Laconte (papers proliferate in France, Wood visits lab and publishes account in Sept. 1904). 1905, publications on N-rays mostly disappear. 1906, he refuses more controlled test for N-rays

Response of Scientific Community to N-Rays

Robert Wood travels to Nancy. Had failed to replicate studies. Wondered what was going on in the lab. Found inadequate methods. Found results matched what people expected to see (even when Wood messed up the experimental setup).

Blondot’s Case

Not fraud: thought he was really detecting N-rays. No data fabrication/falsification or plagiarism. Recorded what he actually saw (or thought).

Didier Raoult and Hydroxychloroquine

Iconoclast personality

Believed a combination of hydroxychloroquine and azithromycin could cure COVID-19 because it was effective for other diseases (like Q-fever)

Cut research/experiment short (from two weeks to six days because he believed he saw results - viral load v. clinical outcome)

Picked up by media and elected officials

Doctors prescribed it (drug storage result, FDA issued ER-use authorization to access national stockpiles)

One person died after dropping out

Larger RCTs showed no benefit, death rates the same

Raoult’s paper retracted

Drug Test Basics

In vivo, In vitro, animal testing, human testing

In Vivo

In live bodies (animals or humans)

In Vitro

In petri dishes (glass) with cells

Animal Testing

Safety and efficacy

Applicability to humans not always accurate