Munich Center for Mathematical Philosophy (MCMP)

Breadcrumb Navigation


Explanation and Reduction in the Sciences (The Third Jerusalem-MCMP Workshop in the Philosophy of Science)

Jerusalem, 8.-10. February 2018

Idea and Motivation

What is a scientific explanation? Which theories of scientific explanation should we embrace? What does it mean to say that some scientific explanations are reductive? To what extent do some scientific explanations support particular versions ontological reductionism or physicalism? The participants of this workshop will address these questions in the context of explanatory practices in cognitive science and physics.



Day 1 (Thursday, 8 February 2018)

09:30 - 09:50 Gathering and Refreshments
09:50 - 10:00 Opening
10:00 - 11:15 Ophelia Deroy: From Explanatory Gap to Explanatory Division in Cognitive Neuroscience
11:15 - 12:30 Oron Shagrir and Lotem Elber-Dorozko: Computation and the Mechanistic Hierarchy
12:30 - 14:00 Lunch Break
14:00 - 15:15 Orly Shenker: Flat Physicalism
15:15 - 16:30 Stephan Hartmann: Hawking Radiation and Analogue Experiments: A Bayesian Analysis
16:30 - 16:45 Coffee Break
16:45 - 17:30 Reuben Stern: Antireductionist Interventionism
19:00 Conference Dinner

Day 2 (Friday, 9 February 2018)

09:00 - 10:15 Mario Günther: Interventionist Mental Causation and the Methods of Cognitive Neuroscience: The Explanatory Merit of Functional Reduction
10:15 - 11:30 Arnon Levy: Must the Best Explanation Be True?
11:30 - 11:45 Coffee Break
11:45 - 13:00 Ori Hacohen: Mental Representations in Cognitive Explanations
13:00 - 13:15 Concluding remarks

Day 3 (Saturday, 10 February 2018)

Day trip for participants


From Explanatory Gap to Explanatory Division in Cognitive Neuroscience
Ophelia Deroy (LMU Munich)

One of the traditional goals of cognitive neurosciences is framed as explaining how neural structures support diverse cognitive functions. This, in turn, supposes that we can start with a determined set of cognitive functions, which most often draw heavily on historical or commonsensical psychological models (grounding problem) or present structures which are difficult to match to neural circuits (structure problem). While many recommend bridging this explanatory gap by revising our cognitive ontologies (e.g. Price and Friston 2005, Klein 2012, and Poldrack and Yarkoni 2016), I argue that we should reconsider the goals of cognitive neuroscience, and split them into two separate ones: One, which is to match computational descriptions to neural circuits, the other, which is to explain how cognitive systems represent computational processes, in order to monitor

Interventionist Mental Causation and the Methods of Cognitive Neuroscience: The Explanatory Merit of Functional Reduction
Mario Günther (LMU Munich)

We investigate, within Woodward’s (2005) framework of interventionism, to what extent we are justified to derive causal relations between mental states and brain states from methods used in cognitive neuroscience. Assuming a minimal supervenience relation between mental states and brain states, Baumgartner’s (2009) argument excludes that causal relations can be derived from studies of cognitive neuroscience which manipulate mental states in order to measure changes in neural activity. In contrast, it turns out that we are justified in deriving the other direction from brain stimulation studies, i.e. causal relations from brain states to mental states. We show that a theory which functionally reduces mental states to brain states escapes Baumgartner's causal exclusion argument. Those results fit the current interpretations and explanations in cognitive neuroscience remarkably

Mental Representations in Cognitive Explanations
Ori Hacohen (The Hebrew University of Jerusalem)

Many explanations in cognitive science will explain a cognitive phenomenon by appealing to an internal mechanism that involves the manipulation of content-bearing vehicles, called mental representations. Realists about mental representations assume that the success of such explanations hinges on the existence of these internal vehicles of content. They must therefore define what it means for a physical entity to have a specific content, and especially how that physical entity can have content naturally and intrinsically, without it being derived from the intentions of others. Naturalistic theories of content have long aimed at achieving this task, with very limited success so far. I discuss an alternative view of mental representations in cognitive science. On this view, the success of representational explanations is not defined by a natural existence of internal representations, quite the opposite – the internal representations are defined by a successful explanation. Mental representations therefore do not have natural intrinsic content, instead their content is derived from the intentions of humans, not unlike the contents of everyday representations such as maps or stop signs. This will also offer an alternative route to the possible physical reduction of intentionality, or “content-bearing”. While naturalistic theories of content look to reduce “content-bearing” to the properties of the content-bearing vehicle or the system it is a part of, the alternative view will shift the focus towards external cognitive agents and the (physical) processes through which they assign

Hawking Radiation and Analogue Experiments: A Bayesian Analysis
Stephan Hartmann (LMU Munich)

We present a Bayesian analysis of the epistemology of analogue experiments with particular reference to Hawking radiation. Provided such experiments can be 'externally validated' via universality arguments, we prove that they are confirmatory in Bayesian terms. We then provide a formal model for the scaling behaviour of the confirmation measure for multiple distinct realisations of the analogue system and isolate a generic saturation feature. Finally, we demonstrate that different potential analogue realisations could provide different levels of confirmation. Our results thus provide a basis both to formalise the epistemic value of analogue experiments that have been conducted and to advise scientists as to the respective epistemic value of future analogue experiments. The talk is based on joint work with R. Dardashti, K. Thebault and E.

Must the Best Explanation Be True?
Arnon Levy (The Hebrew University of Jerusalem)

My topic, as the title suggests, is whether we should hold that the best scientific explanation of a given phenomenon must also be a true description of the factors that underlie (or lead up to) the phenomenon. In a nutshell, I will argue that the answer is ‘no’. More specifically, I’ll argue that there are no good arguments for a ‘yes’ answer, and given that many successful scientific explanations are idealized, hence untrue, we should accept that the best explanation need not be true. In this vein, I look at three sorts of arguments: an argument from general norms governing descriptions; an argument from the aims of science; and two related arguments connected with scientific realism. All, I suggest, do not support a truth requirement on explanation. I will end with some comments on how we should think of explanation sans a truth requirement. I’ll suggest that what underlies the thought that explanations is the idea that assessments of explanatory power ought to be objective. That goal, I propose, can be met with an account that grounds explanation in understanding rather than truth.

Computation and the Mechanistic Hierarchy
Oron Shagrir and Lotem Elber-Dorozko (The Hebrew University of Jerusalem)

According to the mechanistic view, scientific explanations describe relevant causal structures. In Neuro-cognitive science there is a hierarchy of mechanistic explanations - every component in a mechanism is itself explained by appeal to its underlying causal mechanism. From this standpoint computational models are only explanatory when the variables and relations they describe map onto components and their causal relations (i.e., the mechanism). Further, every computational-mechanistic explanation needs implementation details to ascertain that it describes the actual causal structure.

We raise two challenges to the view that computational explanations are mechanistic explanations. One is that it is not clear how computation can be both ‘medium (i.e., implementation) independent’ and include implementation details, as required by the mechanistic view. The other challenge is that because the relation between computation and mechanism is not a component-whole relation, it is not evident how to place computation in the mechanistic hierarchy of explanations.

Flat Physicalism
Orly Shenker (The Hebrew University of Jerusalem)

Type identity physicalism is an unpopular view, criticized from a variety of angles. In this talk I will present a new version of this view, which we call “flat physicalism”, and show how it meets several objections often raised against identity theories. This identity theory is informed by recent results in the conceptual foundations of physics, and in particular clarifies the notion of “physical kinds” in light of a conceptual analysis of the paradigmatic case of reducing thermodynamics to statistical mechanics. I show how flat physicalism is compatible with the appearance of multiple realization in the special sciences, and (if time permits) present the flat physicalist theory of the mental, explaining how the mental can be anomalous even within a flat type physicalist

Antireductionist Interventionism
Reuben Stern (with Benjamin Eva) (LMU Munich)

Gebharter (2017) has recently used interventionist graphical causal models to argue against the causal efficacy and explanatory relevance of macro-level properties. His argument is roughly that if we assume the axioms of the graphical approach to causal modeling in contexts where variables are allowed to enter into asymmetric supervenience relations, then, as things turn out, macro-level properties are causally inert. Though we find no fault with Gebharter’s application of the formal apparatus of causal graphs, we argue that Gebharter’s conclusion rests on tenuous assumptions about what variable sets are appropriate for causal inference in multi-level contexts, and about the relationship between the presence (or absence) of a directed edge in a causal graph and the metaphysics of causation. We then use causal graphs to develop our own novel approach to assessing causal efficacy in multi-level contexts, and argue that the antireductionist program survives

Practical Information


The Hebrew University of Jerusalem (Edelstein Center for History and Philosophy of Science, Technology and Medicine)


The conference is supported by The Sydney M. Edelstein Center for History and Philosophy of Science, Technology and Medicinethe Alexander von Humboldt Foundation through an Alexander von Humboldt Professorship, the Graduate School of Systemic Neurosciences (GSN) and the Munich Center for Neurosciences (MCN).