Week 3 Study Guide: Making Robust Knowledge & Science in Society

Reading: Kuhn, The Structure of Scientific Revolutions

Core question: What makes scientific knowledge "robust" (solid, trustworthy)?

Three philosophies of science

1. Logical positivism (1920s–60s): Science starts from pure empirical data; theory is just data organized by logic. Theories get "verified" by more data (induction).

   • Problem: You can never prove a generalization true just by piling up confirming cases — the classic "black swan problem" (seeing 1,000 white swans never proves "all swans are white"; one black swan refutes it).

2. Falsificationism (Karl Popper, 1930s–70s): Since theories can't be verified, a good theory is one that could in principle be falsified — and hasn't been yet. Science progresses by bold conjectures and ruthless attempts to refute them, not by confirming what we already believe.

   • Problem: Real scientists don't actually abandon theories the moment one contrary data point appears. Max Planck's quip: science doesn't progress because opponents change their minds — it progresses because they eventually die and a new generation takes over.

3. Underdetermination: Any single hypothesis is never tested alone; it's tested along with a whole web of background assumptions (the Duhem-Quine thesis). If an experiment fails, you don't know which piece of the web to blame — the theory itself, or an auxiliary assumption. This leads to:

   • Underdetermination of theory by data: the data alone are never enough to force you to pick one theory over its rivals — multiple theories can fit the same data. The real problem in science is not "what do the facts say" but "how do we choose between theories that all fit the facts."
   • Related idea: theory-dependence (theory-ladenness) of observation — what counts as "data" already reflects theoretical assumptions; scientists don't approach data with a blank, neutral mind.

Concepts

• Paradigm: Kuhn's central and famously slippery term. A paradigm is not just a theory — it's the whole package the scientific community shares:
  • a landmark achievement/exemplar (a model solved problem, like Newton's Principia) that defines what "good" work looks like
  • shared theories, laws, and concepts
  • shared instruments and techniques
  • shared standards for what counts as a legitimate question and a legitimate solution
  • In short: a paradigm is both a worldview and a form of scientific life/practice.
• Normal science: If you fail, the assumption is you did something wrong, not that the paradigm is broken. Normal science is taught to students through training on the shared exemplars.
• Revolutionary science: Anomalies (persistent unsolved puzzles) accumulate until they trigger a crisis. A new paradigm is proposed; eventually the community converts to it.
• Incommensurability: Rival paradigms are hard to compare directly because:
  • they disagree on which problems even need solving (different standards for what counts as a solved problem)
  • key terms change meaning across paradigms (e.g., "mass" in Newton vs. Einstein; "earth" as fixed vs. moving)
  • practitioners of different paradigms, in a sense, "live in different worlds" — they don't see the same things as significant when looking at the same phenomena
  • Because of this, paradigm choice can't be settled by pure logic/proof alone. Conversion happens through persuasion, helped along by demonstrated problem-solving success, but also by aesthetic appeal (elegance and simplicity) and even generational turnover (as younger scientists gradually take over).

Takeaways: Look beyond the data alone. Knowledge is made through the social, intersubjective work of a community, working within a historical context. We need to look at both the content (theories, data) and the context (community, institutions, history) of how knowledge gets made.

Lecture 7: Science in Society

Core questions: Why does science have authority/trust in some contexts and not others? How has the relationship between science and society/politics changed?

Expertise

• Expertise = skill or knowledge in a domain; it's a source of social power in modern societies, and access to it is usually gatekept through education and credentialing.

• Jasanoff's definition (quoted in lecture): expertise today is really about "judgment in the face of uncertainty," and the ability to exercise that judgment in the public interest — not the possession of settled, certain facts.

• Because expert judgment operates under uncertainty, errors are inevitable over time, and each error risks eroding public trust

• Problem of Legitimation asks: what justifies giving scientific expertise special authority over political/social decisions in the first place?

• Problem of Extension asks: once we grant that legitimacy, how far should it extend — who else's input (lay people, affected communities) should count, and where should the boundary of "who counts as an expert" be drawn?

• Science as autonomous vs. transparent: should science claim authority by staying independent/insulated from politics, or by being open/public and inviting scrutiny?

• Epstein's point about "scientization of politics" / "politicization of science": increasingly, political and social controversies get converted into technical disputes so that science can "settle" them — but this means each side recruits its own experts, turning science itself into a political battleground (e.g. climate change in politics).

• "Science as politics by other means": the line between science and society blurs — science doesn't just describe the world, it helps make social order, and its power rests partly on rhetoric and persuasion, not on logic/data alone. (This directly echoes Kuhn's point from Lecture 6 that paradigm change is a matter of persuasion, not pure proof — same idea, now applied to science's relationship with the public.)

• "Socially robust knowledge" isn't about scientific claims being logically/epistemically airtight, it needs two layers to achieve that:

  • internal: how the scientific community itself manages disagreement and shares information
  • external: how the community manages its relationship to the wider public, including whether it respects local expertise

Example: the Aquila earthquake trial

• Giampaolo Giuliani (an amateur/independent researcher at the Istituto Nazionale di Fisica Nucleare) had predicted seismic activity before a deadly 2009 earthquake in L'Aquila, Italy. Government-appointed scientific experts publicly reassured the public that a major quake was unlikely shortly before it struck, killing over 300 people. Several scientists were controversially prosecuted (initially convicted, then largely acquitted on appeal) for the manner of their public risk communication.

The "post-1945 moment"

After WWII, science's relationship to politics and power changed dramatically: the image of scientists was elevated, governments actively funded and directed research (WWII, Cold War), and new collaborative/large-scale ways of organizing science emerged — plus a new "social contract for science" (an implicit deal: society funds and trusts science; science delivers useful, reliable knowledge in return).

Reading: Collins & Pinch, The Golem at Large — "The Science of the Lambs"

The story: After the 1986 Chernobyl disaster, radioactive fallout drenched the Cumbrian fells (a sheep-farming region in England) in an unusually heavy rainstorm. UK government scientists (via MAFF, the agriculture ministry) initially and confidently declared there was no health risk, predicting contamination would decline quickly and disappear within about a week or two. Instead, radioactive caesium levels in sheep kept rising for years. A short sales/movement ban became an indefinite one, eventually restricting around a fifth of the UK's sheep — devastating for farmers whose entire yearly income depends on precisely-timed lamb sales.

What went wrong scientifically: Scientists modeled the upland (peaty, acidic) Cumbrian soil using assumptions from lowland clay soil, where caesium quickly binds/locks up chemically. In upland peaty soil, caesium instead stays chemically mobile and keeps getting re-absorbed by grass roots — meaning the sheep were repeatedly re-contaminated, not just contaminated once. The right data (on caesium behavior in soils) had existed since the 1960s, but had only been read in terms of direct physical exposure risk, not the indirect chemical/biological uptake pathway that mattered here — a nice concrete illustration of theory-dependence of observation: the same old data looked irrelevant until a new question made it relevant.

Why trust broke down:

1. False certainty, not honest uncertainty. Scientists and officials repeatedly declared confident, reassuring timelines ("over within a week or ten days") that then had to be reversed. Farmers didn't mind that scientists made mistakes (farmers are used to an uncertain world) — what damaged trust was the refusal to admit uncertainty or error, and the swagger with which wrong predictions were delivered.
2. Dismissal of local expertise. Farmers had deep practical knowledge — about where water/radiation would pool on their land, about the delicate, flexible timing needed to bring lambs to market at the right condition, about the fact that sheep would never eat imported straw, about how fencing sheep into experimental plots would cause them to "waste." Scientists ignored this knowledge (e.g., ran a fenced bentonite-mineral experiment the farmers said wouldn't work — and it later failed for exactly the reason the farmers gave) and imposed rigid, bureaucratic rules (like "just wait to sell" or "feed them straw") that showed no understanding of how hill farming actually works.
3. Farmers directly witnessing the ordinary messiness of science. Watching scientists struggle to calibrate instruments, get inconsistent readings, and even literally re-measure sheep until fewer "failed" contamination tests, farmers saw firsthand the normal, open-ended, imperfect nature of real ("golem") science — which sharply contradicted the public face of certainty officials presented. (This is Collins & Pinch's core "golem" metaphor from the book as a whole: science is like a golem — powerful but clumsy, human-made, not perfect or infallible, and dangerous mainly when we expect too much certainty from it.)

Farmers ended up oscillating between two equally mistaken views of science — first "science = certainty" (which shattered when predictions failed), then, in overcorrection, "science = pure political conspiracy" serving government interests. Collins & Pinch argue both extremes are wrong. The healthier position is to recognize science as fallible, human, socially embedded work — good, useful, and worth trusting, but never infallible or context-free — and to actively value local expertise alongside institutional science.