Conflict of Interest and the Principle of Total Evidence (with J. Landes), in Piazza, M., Tesi, M. and Vigiani, P. (eds.), Reasoning with Imperfect Information in Social Settings, Pisa: Edizioni della Normale.
Randomized controlled trials (RCTs) are often treated as the gold standard of medical research. Yet, RCTs are also criticized for being subject to biases (e.g., small size, inadequate blinding) that too often make them less than perfectly reliable. In this light, should one consider other sources of evidence? In the Variety-of-Evidence literature, it has been argued that all sources of evidence can in principle be useful, no matter their reliability (Osimani and Landes 2023). Meanwhile, it was observed that most medical trials suffer from conflicts of interest (Roseman et al 2011). Available reviews, however, suggest that conflicts of interest raise the probability of biased estimates but also that studies subject to conflicts of interest are more reliable in virtue of their better design/quality (Lexchin et al 2003). So the question arises, would one here, too, benefit from considering all sources of evidence? We build a Bayesian model to shed light on the matter. We argue that, first, studies subject to conflicts of interest can improve confirmation despite their ambiguous effects and, second, neither quality nor conflict of interest is generally more relevant to confirmation, whence it is unjustified to privilege one over the other.
Constitution, Non-causal Explanation, and Demarcation (with M. Baumgartner), Ergo, forthcoming [preprint]
In philosophy of science, constitutive explanations have attracted much attention since Craver’s influential book Explaining the Brain (2007). His Mutual Manipulability (MM) theory of constitution aimed to explicate constitution as a non-causal explanatory relation and to demarcate between constituents and non-constituents. But MM received decisive criticism. In response, Craver et al. (2021) have recently proposed a new theory, called Matched Interlevel Experiments (MIE), which is currently gaining traction in various fields. The authors claim that MIE retains “the spirit of MM without conceptual confusion”. Our paper argues that this claim is not borne out: neither does MIE meet MM’s objectives nor is it free of conceptual confusion. At the same time, we show that it is possible to meet MM’s objectives in a conceptually sound manner—by adopting the so-called No De-Coupling theory of constitution.
Confirmation by Robustness Analysis: A Bayesian Account (with J. Landes), Erkenntnis, doi: 10.1007/s10670-022-00537-7, forthcoming [journal]
Some authors claim that minimal models have limited epistemic value (Fumagalli, 2016; Grüne-Yanoff, 2009a). Others defend the epistemic benefits of modelling by invoking the role of robustness analysis for hypothesis confirmation (see, e.g., Levins, 1966; Kuorikoski et al., 2010) but such arguments find much resistance (see, e.g., Odenbaugh & Alexandrova, 2011). In this paper, we offer a Bayesian rationalization and defence of the view that robustness analysis can play a confirmatory role, and thereby shed light on the potential of minimal models for hypothesis confirmation. We illustrate our argument by reference to a case study from macroeconomics. At the same time, we also show that there are cases in which robustness analysis is detrimental to confirmation. We characterize these cases and link them to recent investigations on evidential variety (Landes, 2020b, 2021; Osimani and Landes, forthcoming). We conclude that robustness analysis over minimal models can confirm, but its confirmatory value depends on concrete circumstances.
The PC Algorithm and the Inference to Constitution (with M. Baumgartner), The British Journal for the Philosophy of Science, 2023, 74(2), 405-29 [journal]
Alexander Gebharter (2017) has proposed to use one of the best known Bayesian network (BN) causal discovery algorithms, PC, to identify the constitutive dependencies underwriting mechanistic explanations. His proposal assumes that mechanistic constitution behaves like deterministic direct causation, such that PC is directly applicable to mixed variable sets featuring both causal and constitutive dependencies. Gebharter claims that such mixed sets, under certain restrictions, comply with PC’s background assumptions. The aim of this paper is to show that Gebharter’s proposal incurs severe problems, ultimately rooted in the widespread non-compliance of mechanistic systems with PC’s assumptions. This casts severe doubts on the attempt to implicitly define constitution as a form of deterministic direct causation complying with PC’s assumptions.
Variable Definition and Independent Components (with A. Moneta and M. Capasso), Philosophy of Science, 2021, 88(5), 784-95 [journal]
In the causal modelling literature, it is well known that “ill-defined” variables may give rise to “ambiguous manipulations” (Spirtes and Scheines, 2004). Here, we illustrate how ill-defined variables may also induce mis- takes in causal inference when standard causal search methods are applied (Spirtes et al., 2000; Pearl, 2009). To address the problem, we introduce a representation framework, which exploits an independent component representation of the data, and demonstrate its potential for detecting ill-defined variables and avoiding mistaken causal inferences.
Horizontal Surgicality and Mechanistic Constitution (with M. Baumgartner and B. Krickel), Erkenntnis, 2020, 85: 417-30 [journal]
While ideal (surgical) interventions are acknowledged by many as valuable tools for the analysis of causation, recent discussions have shown that, since there are no ideal interventions on upper-level phenomena, which non-reductively supervene on their underlying mechanisms, interventions cannot—contrary to a popular opinion—ground an informative analysis of mechanistic constitution. This has led some to abandon the project of analyzing constitution in interventionist terms. By contrast, this paper defines the notion of a horizontally surgical intervention, and argues that, when combined with some innocuous metaphysical principles about the relation between upper and lower levels of mechanisms, that notion delivers a sufficient condition for constitution. This, in turn, strengthens the case for an interventionist analysis of constitution.
Hypothetical Interventions and Belief Changes (with H. Andreas), Foundations of Science, 2019, 24: 681-704 [journal]
According toWoodward’s (2003) influential account of explanation, explanations have a counterfactual structure, and explanatory counterfactuals are analysed in terms of causal relations and interventions. In this paper, we provide a formal semantics of explanatory counterfactuals based on a Ramsey Test semantics of conditionals. Like Woodward’s account, our account is guided by causal considerations. Unlike Woodward’s account, it makes no reference to causal graphs and it also covers cases of explanation where interventions are impossible.
Alexander Gebharter: Causal Nets, Interventionism, and Mechanisms. Philosophical Foundations and Applications, Journal for General Philosophy of Science, 2018, 49: 481–85 [journal]
Malfunctions and Teleology: On the (Dim) Chances of Statistical Accounts of Functions, European Journal for Philosophy of Science, 2017, 7(2): 319–35 [journal]
The core idea of statistical accounts of biological functions is that to function normally is to provide a statistically typical contribution to some goal state of the organism. In this way, statistical accounts purport to naturalize the teleological notion of function in terms of statistical facts. Boorse’s (1977) original biostatistical account was criticized for failing to distinguish functions from malfunctions. Recently, many have attempted to circumvent the criticism (Boorse, 2014; Kraemer, 2013; Garson and Piccinini, 2014; Hausman, 2012, 2014). Here, I review such attempts and find them inadequate. The reason, ultimately, is that functional attribution depends on how traits would behave in relevant situations, a condition that resists statistical characterizations in terms of how they typically behave. This, I conclude, undermines the attempt to naturalize functions in statistical terms.
An Abductive Theory of Constitution (with M. Baumgartner), Philosophy of Science, 2017, 84(2): 214–33 [journal]
The first part of this article finds Craver’s mutual manipulability theory (MM) of constitution inadequate, as it definitionally ties constitution to the feasibility of ideal experiments, which, however, are unrealizable in principle. As an alternative, the second part develops an abductive theory of constitution (NDC), which exploits the fact that phenomena and their constituents are unbreakably coupled via common causes. The best explanation for this fact is the existence of an additional dependence relation, namely, constitution. NDC has important ramifications for constitutional discovery—most notably, that there is no experimentum crucis for constitution, not even under ideal discovery circumstances.
Can Interventions Rescue Glennan’s Mechanistic Account of Causality? The British Journal for the Philosophy of Science, 2016, 67(4): 1155–83 [journal]
Glennan ([2011]) appeals to interventions to solve the ontological and explanatory regresses that threaten his mechanistic account of causality (Glennan, 1996, 2002). I argue that Glennan’s manoeuvre fails. The appeal to interventions is not able to address the ontological regress, and it blocks the explanatory regress only at the cost of making the account inapplicable to non-modular mechanisms. I offer a solution to the explanatory regress that makes use of dynamic Bayesian networks.Myargument is illustrated by a case study from systems biology, namely, the mechanism for the irreversibility of apoptosis. I conclude by pointing out the implications of my argument for Glennan’s mechanistic account of causality and, more generally, for accounts of mechanistic explanation based on interventions.
How to Model Mechanistic Hierarchies, Philosophy of Science, 2016, 83(5): 946–58 (journal]
Mechanisms are usually viewed as hierarchical, with lower levels of a mechanism influencing, and decomposing, its higher-level behaviour. To draw quantitative predictions from a model of a mechanism, the model must capture this hierarchical aspect. Recursive Bayesian networks (RBNs) were put forward as a means to model mechanistic hierarchies (Casini et al., 2011) by decomposing variables into their constituting causal networks. The proposal was criticized by Gebharter (2014). He proposes an alternative formalism, which decomposes arrows. Here, I defend RBNs from the criticism and argue that they o er a better representation of mechanistic hierarchies than the rival account.
Agent-based Models and Causality: A Methodological Appraisal (with G. Manzo), The IAS Working Paper Series, Linköping University Electronic Press, 2016:7 [open-access]
Computational agent-based models are entering the toolbox of quantitative sociologists. However, markedly contrasting views still exist as to its capacity to contribute to causally-oriented empirical research. Building on selected works across disciplines ranging from computer science to philosophy, we connect scholarship on causality, mechanisms, and simulation methods, and provide the first systematic discussion on how, if at all, computational agent-based models warrant causal inference. First, we argue that this method can produce causally-relevant evidence when (and only when) specific conditions are met. Then, we show that data-driven methods for causal inference face analogous challenges. Finally, upon endorsing a pragmatist view of evidence, we defend an approach to causal analysis that combines evidence from agent-based modeling and data-driven methods. This evidential variety lends credibility to causal inference in virtue of drawing on complementary, and equally important, kinds of evidence.
Not-so-minimal Models. Between Isolation and Imagination, Philosophy of the Social Sciences, 2014, 44(5): 646–72 [journal]
What can we learn from ‘minimal’ economic models? I argue that learning from such models is not limited to conceptual explorations—which show how something could be the case—but may extend to explanations of real economic phenomena—which show how something is the case. A model may be minimal qua certain world-linking properties, and yet ‘not-so-minimal’ qua learning, provided it is externally valid. This in turn depends on using the right principles for model building, and not necessarily ‘isolating’ principles. My argument is buttressed by a case study from computational economics, namely two ABMs of asset pricing.
Causation: Many Words, One Thing?, THEORIA, 2012, 27(74): 203-19 [open-access]
How many notions of cause are there? The causality literature is witnessing a flourishing of pluralist positions. Here I focus on a recent debate on whether interpreting causality in terms of inferential relations commits one to semantic pluralism (Reiss 2011) or not (Williamson 2006). I argue that inferentialism is compatible with a ‘weak’ form of monism, where causality is envisaged as one, vague cluster concept. I offer two arguments for this, one for vagueness, one for uniqueness. Finally, I qualify in what sense the resulting form of monism is ‘weak’.
Models for Prediction, Explanation and Control: Recursive Bayesian Networks (with P.M. Illari, F. Russo, and J. Williamson), THEORIA, 2011, 26(70): 5–33 [open-access]
The Recursive Bayesian Net (RBN) formalism was originally developed for modelling nested causal relationships. In this paper we argue that the formalism can also be applied to modelling the hierarchical structure of mechanisms. The resulting network contains quantitative information about probabilities, as well as qualitative information about mechanistic structure and causal relations. Since information about probabilities, mechanisms and causal relations is vital for prediction, explanation and control respectively, an RBN can be applied to all these tasks. We show in particular how a simple two-level RBN can be used to model a mechanism in cancer science. The higher level of our model contains variables at the clinical level, while the lower level maps the structure of the cell’s mechanism for apoptosis.