Course Overview

The focus in this course is on the errors humans tend to make, and the approaches science methodology has given us (and we are still developing) to prevent or at least minimize those errors. Learn more »

    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • Facts vs. Values
          • a. Facts: Objectively true claims about reality. Everything that is the case. What is, descriptively, including spatial relations, causal relations, attributes of objects, etc.
          • b. Values: What is of value, important, of worth. Oughts, shoulds, etcs.
        • Democracy vs. Epistocracy
          • a. Democracy: A system of government wherein a society’s citizens have more or less equal input into policies.
          • b. Epistocracy: A system of government wherein a particular subset of a society—privileged by their education or other markers of expertise—decides policies.
        • Scientific expertise has utility for political decision-making.  
        • Social and behavioral aspects of the world can be approached scientifically and, therefore, have relevant experts.  
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • An instrument can have greater precision and accuracy than direct observation or the instruments used to test and calibrate it.  
        • Validity of a Measure: Is the measure yielding information about the target entity in out in the world, given that the target is something real (i.e., concept is valid).  
        • Challenges in validating the use of an instrument:  
        • Techniques for validating instruments:  
          • a. Interactive exploration: Testing an instrument by changing the thing it is measuring in ways you know through other means, and seeing if the instrument recognizes the changes appropriately (e.g. does driving increase a car's odometer; see how singing higher and lower notes affects a sound spectrograph; sprayable electrons in Hacking reading).  
          • b. Comparison of multiple instruments (e.g. thermometers)  
          • c. Comparison to direct observation (e.g. naked sight compared to sight with a magnifying glass)  
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • Statistical Uncertainty/Error: Differences between reality and our measurement on the basis of random imprecisions.
          • a. All measurements have a certain amount of variance, which are just differences between multiple measurements due to error and/or genuine variation in the sample. These differences will not all go in the same direction.  
          • b. Statistical uncertainty can be reduced by averaging a larger amount of data.  
        • Systematic Uncertainty/Error: Differences between reality and our measurement that skew our results in one direction.
          • a. Such measurements will show a consistent bias, that is, a consistent deviation from reality in one direction.  
          • b. Systematic uncertainty cannot be reduced by averaging a larger amount of data.  
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • Scientific Optimism: An attitude of optimism that persistence and iteration on a difficult scientific problem will eventually pay off with interesting insights into your problem.  
        • Skeptical/Gatekeeping Function: Science is in the business of rigorously testing claims against experience, rather than merely accepting them.  
        • Discovery/Innovation Function: Science is in the business of generating new theories for how to explain the world. This is both difficult (requires resources, uncertain success) and important (need to make decisions, wouldn’t have anything to “gatekeep” if new scientific ideas weren’t being generated). 
        • Omnivorous Science: Constantly learning new techniques, exposure to a variety of hypotheses & theories, interdisciplinary discussion, etc.  Important to progress because there are payoffs for learning novel experimental/technological/theoretical techniques and questions/problems from many domains of science, even beyond the one that one starts from. 
        • Zero Sum Games vs. Enlarging the Pie: A can-do spirit goes along with optimism that problems can be solved by enlarging the pie, not just redistributing zero-sum goods.
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • Correlation is insufficient to demonstrate causation because there are other causal structures that lead to correlation: 
        • Randomized Controlled Trial (RCT): An attempt to identify causal relations by randomly assigning subjects into two groups and then performing an experimental intervention on the subjects in one of the groups.   
          • a. Experimental Intervention: The act of an experimenter changing one variable in a possible causal network. 
          • b. Randomized Assignment: Given sufficient sample size, randomized assignment rules out confounds by  distributing variation randomly between the two groups, thereby avoiding systematic differences except as the result of the intended intervention. 
          • c. Control Condition: Comparison of an experimental to a control condition is necessary in order to distinguish effect of intervention from changes that would have occurred without the intervention.  
          • d. Sampling: A study of a well-chosen sample can tell you something about the population (through induction), especially if it was selected in such a way as to avoid any systematic differences between the sample and the rest of the population. It is often difficult or even impossible to capture a perfectly representative sample, so scientists do the best they can. For example, many psychology studies are done with college students because they are accessible, but such samples differ systematically from the general population. Inferences from samples to a larger population need to take such differences into account.
        • Causation: X causes Y if and only if X and Y are correlated under interventions on X.   
          • a. This is a technical notion, which overlaps with but is slightly different from everyday usage. For example, everyday usage of the word “cause” can be influenced by moral considerations, the complexity of a causal mechanism, and/or the nature of the mechanism. We typically don’t say that the big bang "caused" this sequence of letters, or that the presence of oxygen caused the forest fire, etc, but scientifically they are part of the causal history of those phenomena. 
          • b. There can be other evidence for causation, even when actually performing an intervention is not feasible. However, saying something is a cause implies that there is in principle a relationship under an intervention.  
      • C. CONCEPT APPLICATION
    • LEARNING GOALS  

      • A. ATTITUDES
        • Appreciate that we can sometimes get very good evidence for a causal hypothesis, even in the absence of decisive RCTs.  
        • Be wary of potential confounds in apparent evidence for causality.  
      • B. CONCEPT ACQUISITION
        • In many cases it is not possible to conduct a true RCT to test causality, for practical or ethical reasons.
        • There are non-RCT forms of evidence for causal hypotheses, which are less conclusive than RCTs but together can offer strong evidence for causation. These include:
          • a. Prior plausibility: Can a plausible mechanism be constructed, or is there some other basis for interpreting the current evidence in terms of one causal structure over another, such as data from other studies?  
          • b. Temporality/temporal sequence: Did the hypothesized cause precede the effect? 
          • c. Dose-response curve: Do the quantities of the hypothesized cause correlate with the quantity, severity, or frequency of the hypothesized effect across ranges?  
          • d. Consistency across contexts: Does the correlation appear across diverse contexts? 
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • Multiple Causation: Any given effect may be brought about by a complex combination of many causes (which may interact with each other), with varying degrees of influence on the outcome. (see Topic XIII).  
        • Singular Causation: A causal relation between specific events — i.e., Event A caused Event B.  
        • General Causation: A causal relation between variables — i.e., X causes Y.  
        • Causation as Production: There is a spatiotemporally connected series of causal connections between two events or event types (i.e., the kind of causation people have in mind when they say there is no action at a distance; e.g., commission).  
        • Causation as Dependence: If X hadn’t happened, Y would not have happened (counterfactual dependence, e.g. omission, double prevention [prevention of a prevention]). 
        • Decision-making involves not only assessment of the outcome, but also the agent’s causal role in the production of the outcome (i.e., omission vs. commission, e.g. trolley dilemma).  
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • Signal: Aspects of observations or stimuli that provide useful information about the target of interest, as opposed to noise.  
        • Noise: The aspects of observations or stimuli that distract from, dilute, or get confused with signal, and are not signal (i.e., do not provide useful information about the target of interest).   
        • Observations/stimuli subject to confusion between signal and noise include communication, measurements, descriptions, etc.   
        • Signal-to-Noise Ratio: The relative strength of signal compared to the relative strength of noise in a given context. Obtaining meaningful information from the world requires distinguishing signal from noise. Therefore, human cognition (both scientific and otherwise) relies on techniques and tools to suppress noise and/or amplify signal (i.e., increase signal-to-noise ratio).    
        • It is possible to design filters to increase the signal-to-noise ratio, if you know where the noise is going to appear.  
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • People are (evolutionarily?) disposed to over-perceive signal (i.e., noise often gets misinterpreted as signal), perhaps because the cost of missing real signal (false negatives) is typically higher than the cost of mistaking noise for signal (false positives).   
        • People tend to see any regularity as a pattern (i.e., see more signal than there is), even when “patterns” occur by chance (i.e. are pure noise), e.g.: People underestimate the frequency of apparent patterns produced by randomness, leading to overperception of spurious signal much more frequently than people account for.  (Events that are just coincidental are much more likely than most people expect.)  
        • Gambler’s fallacy: Expecting that streaks will be broken, such that future results will “average out” earlier ones, even when all trials are independent. 
        • Hot-hand fallacy: Expecting that streaks will continue, even when all trials are independent.  
        • Look Elsewhere Effect: Even if there is a low probability of pure noise passing a given threshold for signal, if we look at enough noise some of it will pass that threshold by chance. I.e., if there is a low probability of obtaining a false positive in any given instance, the more times you try (the more questions you ask, measures you take, or studies you run without statistical correction), the more you increase the probability of getting a false positive.  This occurs when one:
        • Statistical Significance: How unlikely a given set of results would be if the null hypothesis were true (i.e. if the hypothesized effect did not actually exist).  
        • [Technical term: P-values: the probability of getting a result as extreme or more if in fact the hypothesis is false, simply through random noise.]  
      • C. CONCEPT APPLICATION
    • LEARNING GOALS  

      • A. ATTITUDES
        • Given some degree of uncertainty, appreciate that different kinds of errors come with different costs, such that in some cases it is worthwhile to presume the less likely alternative because the error you risk is less costly. 
      • B. CONCEPT ACQUISITION
        • False Positive/Type I Errors: A test yields a positive result, but in fact the condition is not present.  
        • False Negative/Type II Errors: A test yields a negative result, but in fact the condition is present.  
        • There is always the possibility of a trade-off—for a given test, one can reduce the risk of false positives by increasing the risk of false negatives, and vice-versa.   
        • Good decision-making under uncertainty involves having sufficient signal (an adequate test) and setting your threshold appropriately for the relative costs of false positives and false negatives.     
        • In some classification cases like pornography identification or graduate school admissions, there may not be a “truth of the matter” so there aren’t true “false” positives or “true” negatives, although a threshold must still be set.  
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
        • Recognize that every proposition comes with a degree of uncertainty.  
          • e.g., not be impressed by statements made with 100% confidence, no error bars or confidence intervals on claimed measurements, seeking definitive answers when only probabilistic information is available, not recognizing that probabilistic information is better than no information.
        • Value and defend scientific expressions of uncertainty.   
      • B. CONCEPT ACQUISITION
        • Credence: level of confidence that a claim is true, from 0 to 1.    
        • Confidence: essentially a synonym for credence, as in “level of confidence,” instead of colloquial meaning, “state of having a lot of confidence.”  
        • Accuracy: How frequently one is correct; proximity to a true value.  
        • Calibration: How closely confidence and accuracy correspond; that is, how accurate a person or system is at estimating the probability that they are correct.  
        • Because every proposition comes with a degree of uncertainty:
          • a. Partial and probabilistic information still has value.  
          • b. Back-up plans are important because no information is absolutely certain.  
          • c. It is important to invest in calibrating where you are more and less likely to be right, as opposed to being overinvested in being “right.”  
          • d. Scientific culture primarily uses a language of probabilities, not certain facts.  
          • e. Even correctly-done science will obtain incorrect results some of the time.  
      • B. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • Confidence Interval: A range within which a true value of interest lies with a specified probability.  
          • Most commonly a 95% confidence interval, which means there is a 5% chance the true value lies outside the range specified.
        • Error Bars: Smaller bars on a graph that show the range of likely true values around the observed value, typically a 95% confidence interval, or the observed value +/- the standard error or standard deviation. 
        • Scientific culture at its best reinforces the importance of uncertainty by offering respect and career advancement to people on the basis of calibration as well as accuracy. In attaching the ego to calibration as well as accuracy, this discourages scientists from being overly attached to their ideas being “right,” encouraging them to prioritize truth over having been right. 
        • People (including many experts) tend to over-estimate their accuracy at high confidence levels (and under-estimate it at low-confidence levels).     
        • People often use a source’s confidence as a cue to credibility, but appropriately discount confidence when they have evidence of poor calibration.   
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • Multiple Causation: Any given effect may be brought about by a complex combination of many causes (which may interact with each other), with varying degrees of influence on the outcome.  
        • Orders of Understanding: When there are multiple causes of a given outcome, it is often the case that some causes are much more impactful than others. In these cases, we draw a rough qualitative distinction between the cause(s) with the greatest impact for a given effect (first order cause/explanation), the causes with somewhat less impact (second order), and the much less influential causes (third and higher order).  
        • Effect Size: The size of the effect under examination. (e.g. how much being overweight affects health is the effect size of obesity on health). 
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
        • Value processes of discovering and correcting errors, recognizing that formal statistics and logic are needed to counteract various types of (frequently unrecognized) errors, despite high confidence. Be aware of the pitfalls of human reasoning.  
      • B. CONCEPT ACQUISITION
        • Availability heuristic: Cases in which people use how readily something comes to mind as a proxy for an estimate of its probability. 
        • Representativeness heuristic: Cases in which how representative something is of a category or outcome is used as a proxy /for evaluating how likely the category membership or outcome is (not taking base rates into account). 
        • Anchoring heuristic: Cases in which an estimate is made by anchoring on a number that’s provided and potentially irrelevant and adjusting, typically insufficiently.   
        • Base rate neglect: People frequently overlook the importance of base rates when calculating the probability of an event based on probabilities that seem more relevant to the specific case.   
        • Base rates: The base frequency of a given attribute in a whole population.   
          • a. Bayesian reasoning: Systematically combining information regarding new evidence with prior beliefs to determine the probability of a hypothesis.
          • b. Bayes Rule: $$P(A|B) = (P(B|A)*P(A))/P(B)$$ where $$A$$ & $$B$$ are events and $$P(B)$$ ≠ $$0$$.
            • The formal statistical rule for applying Bayesian reasoning. When updating beliefs, final credence level should be influenced by initial credence level and the strength of the new evidence.
        • Peak-End Rule: The tendency to remember the peak, or highlight, and the very end of an experience, and to take them as more representative of the experience as a whole than they really are.
      • C. CONCEPT APPLICATION
        • Recognize and resist instances of the availability heuristic in everyday and scientific contexts.   
        • Recognize and resist instances of the representativeness heuristic in everyday and scientific contexts.  
        • Recognize and resist instances of the anchoring heuristic in everyday and scientific contexts.   
        • Recognize and resist instances of base rate neglect in everyday and scientific contexts.  
        • Given an example in which a person updates her belief, identify the two factors that should influence her final credence level (initial credence level and strength of the evidence), and recognize that Bayes rule provides a formal specification of how to do so. 
    • LEARNING GOALS

    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • The boundaries demarcating science from non-science and distinguishing among the categories of pathological science, pseudo-science, fraudulent science, poorly-done science, and good science can often be difficult, with overlapping and fuzzy boundaries between categories. 
        • Pathological Science Indicators:
          • a. The effect is produced by a barely detectable cause, and the magnitude of the effect is substantially independent of the intensity of the cause.  
          • b. The effect is barely detectable, or has very low statistical significance. Claims of great accuracy. 
          • c. Involving fantastic theories contrary to experience. 
          • d. Criticisms are met with ad hoc excuses. 
          • e. Ratio of supporters to critics rises to near 50%, then drops back to near zero.  
          • f. Conclusion-motivated design & analysis. 
          • g. Pseudo-science is characterized by using scientific vocabulary without aligning with the corresponding concepts or engaging in real scientific practices (i.e.,  science being “skin deep,” not scientific below the surface). 
        • Fraudulent science involves intentional deception, such as deliberately fabricating data or deliberately deceiving the reader about the strength of evidence.  
        • Poorly-done science, e.g. failure to consider confounds, failure to use best practices in terms of data collection and analysis (e.g., small sample size, look elsewhere effect).  
        • Unintentional self-deception can be involved in justifying poor practices and/or interpretations in pathological, pseudo-, & poorly-done science. 
        • Motivation to support a particular conclusion (i.e., science undertaken to support a given conclusion, rather than to discover the truth) can be a feature of poorly done or pathological science.  
        • Good science:
          • a. Will get the wrong answer some of the time, e.g., via statistical flukes.  
          • b. Entails good faith engagement with the alternative hypotheses through a search for evidence that you are wrong.  
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
      • B. CONCEPT ACQUISITION
        • Confirmation bias: Seeking or otherwise favoring evidence consistent with what is already believed or what is being tested. 
          • a. Selective exposure: Selectively seeking or exposing oneself to evidence that is likely to conform to prior beliefs or working hypotheses.
          • b. Biased assimilation: Systematically favoring or discounting evidence to render that evidence more compatible with pre-existing beliefs or working hypotheses.
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
        • One should always be looking for ways that we get things wrong (by fooling ourselves or due to bugs in our reasoning processes) so that we can invent better procedures. 
      • B. CONCEPT ACQUISITION
        • Blind analysis: Making all decisions regarding data analysis before the results of interest are unveiled, such that expectations about the results do not bias the analysis. Usually co-occurs with a commitment to publicize the results however they turn out. 
        • Examples of analysis decisions for which blind analysis could be useful: 
        • Confirmation bias drives the need for blind analysis.  
        • Confirmation bias is pervasive and doesn’t necessarily indicate any fraudulent activity.  
        • Approaches to reducing confirmation bias other than blind analysis:
          • a. Preregistration: A research group publicly commits to a specific set of methods and analyses before they conduct their research. 
          • b. Registered replication: [A] research group(s) commits to a specific set of methods and procedures to verify the result of an earlier work (typically with the input of the original research team). Results are publicized regardless of outcome. 
          • c. Adversarial collaboration: Scientists with opposing views agree to all the details of how data should be gathered and analyzed before any of the results are known. 
          • d. Peer review: New results are evaluated by other experts in the same field to determine whether they are valid. This only reduces confirmation bias if reviewers don’t share biases. 
        • Scientists are constantly looking for bugs in scientific practices in order to fix them. Blind analysis is just the latest example of scientists recognizing a bug in their practice (e.g., a way of being fooled) and adjusting practice to account for/remove the bug.    
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
        • Not take for granted that consensus offers the best conclusions. 
        • Take seriously (but not as absolute!) the consensus of a group which has reasoned about a question in a careful, appropriate way. 
        • Take seriously (but not as absolute!) the average of a large group's independent estimates of a number, under appropriate conditions. 
      • B. CONCEPT ACQUISITION
        • Wisdom of Crowds: Sometimes groups make better judgments than individuals. This happens when: 
          • a. Judgments are genuinely independent, preventing herd thinking. 
          • b. Members of the group do not share the same biases. 
          • c. There are enough people in the group to balance out random biases or fluctuations (analogous to the need for an adequate sample size). 
          • d. Works especially well when estimating a quantity, where errors may be large but are not systematic. 
        • Herd Thinking: Sometimes groups make worse judgments than individuals. This happens when: 
          • a. Judgments of individuals are influenced by the judgments of others, leading to groupthink and sometimes polarization. 
          • b. Members of the group share biases, which can be exaggerated by discussion and cannot be decreased by averaging judgments. 
        • The enterprise of science is essentially social, and advances in part because scientists look actively for what other scientists might have gotten wrong. This process, including peer review, enables science to iteratively improve.
      • C. CONCEPT APPLICATION
    • LEARNING GOALS
      • A. ATTITUDES
        • Be optimistic that a community can come together to make a decision, even when people begin with heterogeneous values and beliefs.
      • B. CONCEPT ACQUISITION
        • Stakeholders: The set of people who have a stake in the outcome of a decision. This can include people who will implement the decision and all the people affected by it.
        • Experts: The set of people who have the most knowledge/information/expertise about the facts relevant to the decision.
        • Denver Bullet Study: An experiment in group deliberation in which a community came together to share values and knowledge to make a decision about what kind of bullet the Denver Police should use, which had enough stopping power to keep cops safe but was not so harmful as to cause unnecessary damage to citizens (as did classic hollow bullets).
      • C. CONCEPT APPLICATION
    • LEARNING GOALS

      • A. ATTITUDES
        • Avoid assuming the future will continue in one direction, and ready to consider a variety of possible futures when planning. 
      • B. CONCEPT ACQUISITION
        • Scenario Planning: A mode of problem solving which involves considering two important and uncertain dimensions along which the future might vary, imagining what each possible quadrant might look like, and considering how decisions made now will affect the likelihood and desirability of each quadrant. 
      • C. CONCEPT APPLICATION
    • LEARNING GOALS