Identify common logical fallacies and cognitive biases that distort scientific claims, including ad hominem, appeals to authority and confirmation bias

What this dot point is asking

NESA wants you to identify common logical fallacies and cognitive biases in scientific reasoning, name them, and explain how they distort claims. This dot point is examined in short-answer and 5-7 mark questions.

The answer

Logical fallacies are flaws in reasoning. Cognitive biases are systematic mistakes in thinking shaped by how the brain processes information. Both distort scientific debate.

Common logical fallacies

Ad hominem. Attacking the person rather than their argument.

Example: "We can ignore Greta Thunberg on climate because she is a teenager."

Why it is fallacious: a person's identity does not determine whether their argument is correct. Evaluate the evidence.

Appeal to authority. Accepting a claim because an authority figure said it, without evidence.

Example: "Einstein doubted quantum mechanics, so we should too."

Why it is fallacious: even Nobel laureates can be wrong outside their domain of expertise. The evidence is what supports a claim, not the authority of the speaker.

False dichotomy. Presenting only two options when more exist.

Example: "Either nuclear power is the answer or we accept climate catastrophe."

Why it is fallacious: many policy options exist. The framing obscures complexity.

Post hoc ergo propter hoc. "After this, therefore because of this." Treating temporal sequence as causal.

Example: "Vaccination at 12 months, autism diagnosed at 24 months, therefore vaccination caused autism." The two events both occur after birth but the temporal sequence alone does not establish causation.

Strawman. Attacking a misrepresented version of an opponent's argument.

Example: "Climate scientists want to ban all cars." Climate scientists generally do not advocate this; the strawman is easier to attack than their actual policy proposals.

Slippery slope. Claiming one step inevitably leads to extreme consequences without evidence.

Example: "If we allow GM crops, we will be eating Frankenstein food in ten years."

Naturalistic fallacy. Treating natural as good.

Example: "Vaccines contain unnatural chemicals." The "natural" status of a substance does not determine its safety or effectiveness.

Genetic fallacy. Judging a claim by its origin rather than its merits.

Example: "That research was funded by a pharmaceutical company, so it must be wrong." Funding can introduce bias but does not automatically invalidate the science. Evaluate the methodology.

Argument from ignorance. Treating absence of evidence as evidence of absence (or presence).

Example: "We cannot prove there is no link between Wi-Fi and cancer, so there must be one." The burden of proof rests with the claim. Without evidence, no claim is warranted.

Common cognitive biases

Confirmation bias

Seeking and remembering information that confirms existing beliefs while ignoring contrary evidence. The most pervasive cognitive bias in scientific reasoning.

Anchoring

Over-weighting the first piece of information encountered. The initial estimate biases subsequent judgements.

Availability heuristic

Judging probability by ease of recall. Plane crashes get heavy news coverage, so people overestimate the risk of flying compared with driving.

Dunning-Kruger effect

People with limited knowledge of a topic over-estimate their expertise. People with deep knowledge tend to under-estimate it.

Survivorship bias

Drawing conclusions only from successful examples (the survivors) while ignoring the failures. "Successful companies all do X" ignores all the failed companies that also did X.

Hindsight bias

Believing past events were more predictable than they actually were, after they have happened.

Sunk cost fallacy

Continuing an investment because of past commitment rather than future expected return. Researchers may persist with a failing hypothesis because of years invested.

How biases distort science

Hypothesis formation

Confirmation bias narrows the questions asked.

Study design

Researchers may design experiments more likely to confirm their hypothesis.

Data interpretation

Ambiguous data is interpreted to fit the existing view.

Reporting

Positive results are published; negative ones are filed away (publication bias).

Peer review

Reviewers share field-wide biases and may judge papers favourably that confirm their views.

Mitigations

• Pre-registration of hypotheses and analyses locks in the question before the data is seen.
• Blinding prevents researchers from interpreting data based on group membership.
• Adversarial collaboration pairs researchers with opposing views to design joint experiments.
• Statistical pre-registration of primary endpoints prevents post-hoc redefinition.
• Replication by independent teams with no stake in the original conclusion.

A specific Australian example

The H. pylori case (Module 5). For decades the medical consensus held that stress and acid caused peptic ulcers. Doctors dismissed the bacterial hypothesis as implausible (an instance of consensus bias and ad hominem dismissal of Barry Marshall's outsider status). Marshall countered with self-experimentation, eventually winning the 2005 Nobel Prize. This is an example of consensus bias being overcome by evidence and persistence.