When we seek to understand why some event e occurred, we typically try to identify other events which causally explain e's occurrence. Of course, some causal explanations are better than others. When we ask why Newcastle beat Sunderland, both `because they have better players' and `because they were playing at home' look like acceptable answers, in the sense that they both explain the victory. But Newcastle's having superior players is surely a better explanation. In joint work with Reuben Stern, I've developed a general approach for measuring the strength of causal explanations, which makes crucial use of the theory of causal Bayesian networks. We're now working on applying this framework to provide new perspectives  on questions like `To what extent can mental events provide causal explanations of other mental events and physical events?', and `What is the relationship between the extent to which an agent's actions causally explain an outcome, and the extent to which they are morally responsible for that outcome?'. 

​

I also have more general research interests relating to i.e. the problem of inferring causal structure from empirical data, the ways in which beliefs about causal structure guide reasoning and learning, and the metaphysics of causation. 

 

Related Papers and Talks:

(1) Eva and Stern (2018). Causal Explanatory Power. British Journal for the Philosophy of Science.

(2) Eva, Stern and Hartmann (Forthcoming). The Similarity of Causal Structure. Philosophy of Science. 

(3) Eva and Stern (In Progress). Anti Reductionist Interventionism. [Slides]

(4) Eva and Stern (In Progress). Causal Explanatory Power, Blameworthiness and Negligence. [Slides]

Another central strand of my research focuses on developing connections between Bayesian epistemology and the psychology of reasoning. 

​

In joint work with Stephan Hartmann, i've developed a framework for greatly extending the range of learning experience that can be adequately described in a Bayesian setting, and then applied the framework both to provide new models for understanding actual human reasoning practices and also to solve problems in the normative foundations of Bayesian epistemology and the philosophy of science. Perhaps the most important result of this project so far is that it yields a new understanding of what it means to learn an indicative conditional of the form `if p then q'. This in turn allows us to provide fruitful new models of the way that people evaluate the success of arguments in both scientific and everyday reasoning contexts. I also have a long running interest in applying Bayesian methods to explicate the laws of good scientific reasoning.

​

Related Papers:

(1) Eva, Hartmann and Rafiee Rad (Forthcoming). Learning From Conditionals. Mind. 

(2) Eva and Hartmann (Conditionally Accepted). On the Origins of Old Evidence.  Australasian Journal of Philosophy. 

(3) Singmann, Hartmann and Eva (Forthcoming). A New Probabilistic Explanation of the Modus Ponens--Modus Tollens Asymmetry (with Henrik Singmann and Stephan Hartmann). Proceedings of the 41st Annual Meeting of the Cognitive Science Society. 

(4) Eva and Hartmann (Forthcoming). Special Issue of Synthese on the theme `Reasoning in Physics'. 

(5) Eva and Hartmann (2018). Bayesian Argumentation and Logical Validity. Psychological Review.

(6) Eva and Hartmann (2018). When No Reason For is a Reason Against. Analysis. 

(7) Eva (2017). A-Symmetric Confirmation and Anthropic Skepticism. Synthese.

Much of contemporary epistemology focuses on the analysis of two basic types of epistemic attitude: qualitative belief and numerically graded belief, or `credence'. While I agree that these attitudes undeniably play a central role in human cognition and the articulation of the laws of good reasoning, I also contend that there are situations in which the epistemic states of rational agents are better understood in terms of other types of epistemic attitude. Specifically, I think there is a lot to be gained by reviving the tradition (championed by figures such as Keynes, Koopman and Fine) of conceiving of epistemic states in terms of comparative confidence judgements of the form `I am more confident in p than I am in q' and `I am equally confident in p and q'. 

 

By way of illustration, recall the principle of indifference, which stipulates that when an agent has no evidence concerning which of a set of competing possibilities is the true one, they should distribute their credence equally amongst the possibilities. Famously, it turns out this advice is actually inconsistent. This fact was widely used to motivate the idea that we should conceive of agents' epistemic states in terms of `imprecise credences', and reformulate the principle of indifference as requiring agents to adopt maximally imprecise credences when they have no relevant evidence. But it turns out that this reformulation of the principle also ends up giving agents very questionable advice. In recent work, I consider the possibility of reformulating the principle of indifference as a constraint on comparative confidence judgements. It turns out that there are two natural comparative reformulations and only one of them is plausible (indeed, it solves all of the major issues with existing formulations from the literature.)

​

Related Papers:

(1) Eva (Forthcoming). Principles of Indifference. The Journal of Philosophy

(2) Eva  (Under Review). Comparative Learning. [Draft]

My PhD thesis focused mainly on attempting to clarify the relationship between the actual physical content of quantum mechanics (QM), and the mathematical language in which the theory is standardly formulated (the theory of Hilbert spaces), and applied the resulting insights to shed new light on the metaphysical implications of QM. More specifically, I used category theory and other tools from contemporary mathematical logic to identify the properties that a mathematical language needs to have in order to properly describe the quantum world, and systematically surveyed the philosophical significance of alternative formalisms of QM from the contemporary physics literature. 

More generally, I am now involved in an ongoing project of attempting to use category theory (the most general theory of mathematical structure) to develop a rigorous methodology for cataloguing the characteristic content of scientific theories, and their relationships to other theories. This work promises to yield new insights on the nature of inter-theoretic relationships like reduction, emergence and theoretical equivalence. Relatedly, I am a member of the DFG funded international scientific network on `Category Theory in Philosophy of Science' (2018-2021). 

 

Related Papers:

(1) Dewar and Eva (In Progress). A Categorical Perspective on Symmetry and Equivalence. [Draft]

(2) Eva, Doering and Ozawa (Under Review). A Bridge Between Q-Worlds. [Draft]

(3) Eva (2017). Topos Theoretic Quantum Realism. British Journal for the Philosophy of Science.

(4) Eva (2016). Category Theory and Physical Structuralism. European Journal for the Philosophy of Science.

(5) Eva (2016). Modality and Contextuality in Topos Quantum Theory. Studia Logica.

(6) Eva (2016). Towards a Paraconsistent Quantum Set Theory. Proceedings of the 12th International Workshop

on Quantum Physics and Logic, University of Oxford.

The philosophy of science has been extremely successful in developing rich theoretical frameworks for solving the problem of theory choice, i.e. the problem of choosing between rival scientific theories on the basis of empirical evidence. In contrast, comparatively little progress has been made in developing systematic and rigorous approaches to the problem of theory formation, i.e. the problem of formulating novel scientific concepts and theories in an effective manner. Indeed, many philosophers still agree with Popper that `the act of conceiving or inventing a theory seems...neither to call for logical analysis nor to be susceptible to it' and that `There is no such thing as a logical method of having new ideas, or a logical reconstruction of this process'. I'm interested in trying to overturn this orthodoxy, and attempting to develop concrete, psychologically plausible models of theory formation that can be applied in a wide range of reasoning scenarios.

 

The most promising methodology for this project involves drawing on the rich resources of the cognitive science of creativity in order to identify the heuristics and mechanisms of theory formation at play in human cognition, and then design AI systems which implement those mechanisms. The systems can then be tested and evaluated in concrete reasoning tasks, which will allow us to evaluate the efficacy of the relevant theory formation strategies, and draw normative conclusions concerning the epistemology of theory formation and the nature of good creative reasoning in science. In a sense, this project is trying to take up the gauntlet thrown down by Herbert Simon back in the 70s/80s. Simon argued extensively for the possibility and importance of a logic of theory formation, and also advocated something close to the multi-disciplinary methodology sketched above. Recent advances in AI and `Creative Cognition' make the project both more pressing and more feasible than it was in Simon's time. Check back soon for papers and talks!  

Human like reasoners have both qualitative beliefs (i.e. I believe Newcastle will win avoid relegation from the Premier League) and numerical credences (i.e. I think there's about a 75% chance that Newcastle will avoid relegation from the Premier League). While the rationality norms that govern qualitative belief are typically thought to be logical in character (i.e. my beliefs should be logically consistent and closed under logical entailment), the laws of rational credence are commonly thought to be given by the axioms of probability theory. I've recently spent a lot of time thinking about how these different sets of rationality norms interact with one another, and what that means for our understanding of rationality, learning, suppositional reasoning and the semantics of conditionals. I'm also currently thinking a lot about the epistemic value of qualitative belief for agents equipped more fine grained doxastic attitudes (like numerical credences).  

​

Related Papers:

(1) Eva. The Logic of Conditional Belief. Under Review. [Draft]

(2) Eva, Shear and Fitelson. Four Approaches to Suppositional Judgement. Work in Progress. [Slides]