“Making Scientific Inference More Objective” is an ERC funded research project, which started in September 2015. Our objective is to increase the objectivity of scientific inference. To this end, we calibrate philosophical analysis of scientific objectivity with scientific practice. Please download the project synopsis for a detailed description of the project. Click here for a project abstract.

For more information on the researchers, visit the team page! Below we explain the structure of the project.

**The Conceptual Framework: Scientific Objectivity and Subjective Reasoning.**

What makes a scientific inference objective? Traditional answers are freedom from (non-cognitive) values, absence of bias, intersubjective agreement. But recent philosophical research (e.g., by Helen Longino and Heather Douglas) has challenged these assumptions. We explore to what extent Bayesian inference, which is built on explicitly subjective assumptions (=personal degrees of belief) can support objective findings in science.

**Subproject A: Statistical Inference—Toward a New Logic of Hypothesis Testing**

Null hypothesis significance tests (NHST) are the prevalent form of statistical inference. But how valid is their logic? The lack of clear-cut rules for appraising insignificant test poses major problems to NHST. Moreover, their reject/fail to reject dichotomy fuels publication bias. We rethink the logic of hypothesis tests and address their salient deficits (e.g., by developing a measure of corroboration). We also compare the reliability of Bayesian and frequentist hypothesis tests and explore common grounds between both approaches.

**Subproject B: Causal Inference—Toward a Measure of Causal Strength.**

Determining the strength of causal relations is a central problem for scientific inference. A fruitful framework for doing so is the probabilistic account of causality: causes raise the probabilities of their effects. To avoid spurious correlations, we combine this account with the interventionist framework for causation (=interventions on the cause change the probability of the effect). We compare measures of causal strength from a theoretical and empirical perspective and apply them in various domains, e.g., the epistemology of conditionals and the debate about outcome measures in medicine.

**Subproject C: Explanatory Inference—The Determinants of Explanatory Judgment.**

Inference to the Best Explanation (IBE) is an important mode of inference in many scientific disciplines. But in which sense is IBE objective? And when is it reliable? To address this challenge, we explicate the notion of explanatory power in a Bayesian framework, and we compare it to other determinants of explanatory judgment: (i) a sense of understanding, (ii) scope and simplicity, (iii) logical and statistical relations, (iv) causal salience. We aim at synthesizing these different aspects of explanatory reasoning in order to ground the objectivity of explanatory inference.

**Subproject D: Applications—Statistics, Psychology, Medicine.**

We apply the insights from the previous subprojects to a diversity of problems in science and philosophy: the epistemology of conditionals, evaluating clinical trials in medicine, rethinking the foundations of statistics, and proposals for overcoming the replication crisis in science.

Download the synopsis of the project here. (Note: this is the first part of the actual grant application. Some contents have changed as the project proceeds. The above short description of the subprojects reflects these changes.)