Researching medical education Imperial School of Medicine Faculty Teaching Forum 2 May 2012 Dylan Wiliam

Pasteur’s quadrant

Educational research “An elusive science” (Lagemann)  A search for disciplinary foundations Making social science matter (Flyvbjerg, 2001)  Contrast between analytic rationality and value-rationality  Physical science succeeds when it focuses on analytic rationality  Social science  fails when it focuses on analytic rationality, but  succeeds when it focuses on value-rationality  reasonableness, rather than rationality (Toulmin) as the key criterion

Research methods 101: causality  Does X cause Y?  Given X, Y happened (factual)  Problem: post hoc ergo propter hoc  If X had not happened, Y would not have happened (counterfactual)  Problem: X did happen  So we need to create a parallel world where X did not happen  Same group different time (baseline measurement)  Need to assume stability over time  Different group same time (control group)  Need to assume groups are equivalent  Randomized contolled trial

Plausible rival hypotheses  Example: Smoking cigarettes causes lung cancer  Randomized controlled trial not possible  Have to rely on other methods  Logic of inference-making  Establish the warrant for chosen inferences  Establish that plausible rival interpretations are less warranted

Criteria for causal inferences The environment and disease: association or causation? (Hill, 1967)  Criteria for determining a causal association: 1.strength 2.consistency 3.specificity 4.temporality 5.biological gradient 6.plausibility 7.coherence 8.experimental evidence 9.analogy.

Knowledge Not justified-true-belief Discriminability (Goldman, 1976) Elimination of plausible rival hypotheses Building knowledge involves:  marshalling evidence to support the desired inference  eliminating plausible rival interpretations ‘Plausible’ determined by reference to a theory, a community of practice, or a dominant discourse

Inquiry systems (Churchman, 1971) SystemEvidence LeibnizianRationality LockeanObservation KantianRepresentation HegelianDialectic SingerianValues, ethics and practical consequences

The Lockean inquirer displays the ‘fundamental’ data that all experts agree are accurate and relevant, and then builds a consistent story out of these. The Kantian inquirer displays the same story from different points of view, emphasising thereby that what is put into the story by the internal mode of representation is not given from the outside. But the Hegelian inquirer, using the same data, tells two stories, one supporting the most prominent policy on one side, the other supporting the most promising story on the other side (Churchman, 1971 p. 177). Inquiry systems

The ‘is taken to be’ is a self-imposed imperative of the community. Taken in the context of the whole Singerian theory of inquiry and progress, the imperative has the status of an ethical judgment. That is, the community judges that to accept its instruction is to bring about a suitable tactic or strategy [...]. The acceptance may lead to social actions outside of inquiry, or to new kinds of inquiry, or whatever. Part of the community’s judgement is concerned with the appropriateness of these actions from an ethical point of view. Hence the linguistic puzzle which bothered some empiricists—how the inquiring system can pass linguistically from “is” statements to “ought” statements— is no puzzle at all in the Singerian inquirer: the inquiring system speaks exclusively in the “ought,” the “is” being only a convenient façon de parler when one wants to block out the uncertainty in the discourse. (Churchman, 1971: 202). Singerian inquiry systems

Educational research …can be characterised as a never-ending process of assembling evidence that:  particular inferences are warranted on the basis of the available evidence;  such inferences are more warranted than plausible rival inferences;  the consequences of such inferences are ethically defensible. The basis for warrants, the other plausible interpretations, and the ethical bases for defending the consequences, are themselves constantly open to scrutiny and question.

Effects of feedback Kluger & DeNisi (1996) Review of 3000 research reports Excluding those:  without adequate controls  with poor design  with fewer than 10 participants  where performance was not measured  without details of effect sizes left 131 reports, 607 effect sizes, involving individuals On average feedback does improve performance, but  Effect sizes very different in different studies  38% (50 out of 131) of effect sizes were negative

Getting feedback right is hard Response typeFeedback indicates performance… exceeds goalfalls short of goal Change behaviorExert less effortIncrease effort Change goalIncrease aspirationReduce aspiration Abandon goalDecide goal is too easyDecide goal is too hard Reject feedbackFeedback is ignored

Kinds of feedback (Nyquist, 2003) Weaker feedback only  Knowledge or results (KoR) Feedback only  KoR + clear goals or knowledge of correct results (KCR) Weak formative assessment  KCR+ explanation (KCR+e) Moderate formative assessment  (KCR+e) + specific actions for gap reduction Strong formative assessment  (KCR+e) + activity

Effects of formative assessment (HE) Kind of feedbackCountEffect/sd Weaker feedback only Feedback only Weaker formative assessment Moderate formative assessment Strong formative assessment160.56