Quantcast
Channel: The Brains Blog — Causation
Viewing all articles
Browse latest Browse all 26

Cognitive Science of Philosophy Symposium: Causal Cognition

0
0

Welcome to the Brains Blog’s Symposium series on the Cognitive Science of Philosophy. The aim of the series is to examine the use of empirical methods to generate philosophical insights.

The use of diverse methods in research on causal cognition has come with a plurality of theories about how causal cognition works. However, causal pluralism in psychology and philosophy might be unsatisfactory for various reasons. Where does this leave us as experimental researchers and philosophers engaging with empirical research? In this symposium, Lara Kirfel and Tobias Gerstenberg (Stanford University) argue that we should hold on to a plurality of methods, with David Danks (UCSD) providing commentary.

* * * * * * * * * * * *

Anything goes!
From a pluralism of methods towards a unified theory of causal cognition

Lara Kirfel and Tobias Gerstenberg

“If we want to understand nature, if we want to master our physical surroundings, then we must use all ideas, all methods, and not just a small selection of them.”

Paul Feyerabend, Against Method / Wider den Methodenzwang, 1975
The origins of studying causal cognition

Albert Michotte, a Belgian experimental psychologist working in the post world war era, was one of the first psychologists to study causal cognition. To do so, Michotte invented his own experimental machinery, the “Banc Michotte” (see Figure 1a). In a sophisticated set up consisting of rotating discs with coloured lines, this machine created a visual illusion of moving and interacting physical objects. With the help of this machine, Michotte discovered the famous “launching effect”: if one object stops before another and the other object starts to move directly after, people perceive the first object to have caused the other one to move.

Fig. 1a: Albert Michotte in front of the “Banc Michotte,” ca. 1940. Used with the permission of the KU Leuven Libraries. See also Leyssen 2021.
Fig. 1b: The canonical Michotte collisions as animated video clips, from Bechlivanidis et al. (2019). Reproduced here with the author’s permission.

To study causal perception, psychologists nowadays use computer-generated animations of object collisions (see Figure 1b). In studying under which conditions people perceive causality in the interactions of objects, Michotte not only invented a new experimental paradigm. Research on causal perception has also informed the philosophical question of what causality is. Process theories of causation (Michotte, 1946; Salmon, 1994; Wolff 2007) postulate that there must be a spatiotemporally continuous process (such as a transfer of force) between two objects in order for them to be conceptualized as cause and effect.

A tale of two theories: From causal perception to causal inference

In line with process theories, it sometimes feels like we can directly perceive causation in action. However, at other times we have to infer causality from observing the outcomes of hidden causal processes. Consequently, psychologists developed new experimental paradigms to study how people do so. In a popular paradigm, participants view contingency tables that show how often cause and effect co-occur with one another (see Figure 2a), and are asked if the candidate cause brings about the effect. Because mere observation is not sufficient to distinguish between different candidate causal structures that could have produced the data, researchers in this tradition have also tested how people use interventions to figure out how a causal system works (see Fig 2b). These new experimental paradigms for studying causal learning and inference have shaped various kinds of dependence theories of causation (Lewis, 1973; Pearl, 2000; Woodward, 2007). According to dependence theories, people establish a causal relationship if the outcome is dependent on the cause. Various theories have cashed out differently what dependency means by using formal tools such as probabilities, counterfactuals, or interventions. 

Fig. 2a: Sample contingency table.
Fig. 2b: Causal intervention paradigm, from Rottman & Keil (2012). Reproduced here with the author’s permission.
Breaking the theoretical divide through new methods

Ever since Michotte’s ingenious machine, the toolkit of causality researchers has expanded vastly. While early research paradigms uncovered relevant findings for theories of causal cognition, the diverging methods also led to diverging theories about how causal cognition actually works: Animations of object collisions were often used to study the role of spatio-temporal processes, and interventionist paradigms were used to study the role of dependency relationships between cause and effect. The rivalry between process and dependence accounts over what constitutes causality in people’s minds has long divided the landscape of causal cognition research. In recent years, novel methods for studying causal cognition have started to soften this binary picture. Vignette-based studies and paradigms that rely on people’s intuitive understanding of the physical world suggest that people are sensitive to both information about processes as well as to counterfactual dependence when making causal judgments (Gerstenberg, Goodman, Lagnado & Tenenbaum, 2021). In addition to breaking open the dichotomy between process and dependence accounts, these methods have uncovered new aspects that shape causal cognition. For example, studies on causal judgments often draw on text vignettes describing causal scenarios featuring causal agents or inanimate objects, but vary certain features of the scenario such as the moral properties of the agent’s actions. This research has revealed that an agent’s intentions, and the normative status of their actions influence people’s causal judgments. These findings have given rise to theories highlighting the role of blame attribution in causal judgments.

From a pluralism of methods to a causal pluralism?

In light of the heterogeneous findings stemming from a variety of methods, some philosophers and psychologists have adopted a causal pluralism: The idea that there is no single process underlying causality in people’s minds, but that causal cognition comprises a diverse set of cognitive mechanisms. This methodological pluralism and resulting theoretical pluralism in causal cognition research, however, is somewhat unsatisfactory. First, from a psychological perspective, causal pluralism leads to the problem of determining what the relationship between the different causal concepts is, and how it is determined in what situations what causal concepts apply. Second, from a philosophical perspective, causal pluralism raises the question of what causality really is, and what normative role causality plays in how people interact with the world. 

Where does this leave us as psychologists aiming to study how people think about causality? And where does this leave us as philosophers engaging with the empirical literature on causal cognition? There have been some attempts to bridge divergent findings under the framework of one single theory (Gerstenberg et al., 2021; Lombrozo, 2010), and Dinh and Danks (2021) have recently argued that, in the face of causal pluralism, we should still hold on to a unitary concept of causation in cognition. The results from different experimental paradigms might simply represent different behavioural manifestations of the same cognitive causal process. On such an account, different experimental methods make epistemically accessible different features of the same cognitive mechanism underlying people’s causal cognition.

From a pluralism of methods to a unified theory of causal cognition!

How, then, should we leverage psychology’s broad methodological spectrum in order to probe a unified causal concept? As practitioners working at the intersection of philosophical theory and empirical research, we agree with Dinh and Danks that it is desirable to hold onto a unified view of causation in human cognition. Moreover, we argue that as psychologists and philosophers, we should employ a pluralism of methods precisely because we are interested in a unitary theory of causal cognition. However, we want to make an alternative proposal about what kind of unified picture of causal cognition we should strive for.  While most research in the empirical sciences has focused on uncovering how causal cognition works, less attention has been paid to the question of what causal cognition is for (Woodward, 2021). And while it might be the case that causal cognition draws on a diversity of cognitive processes, it is yet an open question what function it actually serves. On the one hand, the answer to this question might appear obvious, especially when we consider people’s ability to grasp general causal relationships. Understanding general causal relationships enables us to make predictions about the future, inferences about the past, and allows us to choose actions that suit our goals. One the other hand, what function the practice of making causal judgments about particular events plays is less clear. Why do we judge that one cause contributed to an outcome more so than another? Why do we select certain causes in our explanations of events, and dismiss others? Crucially, neither process nor dependence theories fully answer the question of what purpose causal judgments serve.

The function of causal judgments

Fortunately, there are already several candidate proposals about what function causal judgments serve. For example, making causal judgments may help us identify and communicate optimal points of intervention. Causal judgments may also be important for how people learn about the causal structure of the world. Alternatively, we might make causal judgments with the goal of assigning blame or responsibility to causal agents. More proposals are possible here, and the discussion about the function(s) of causal judgment has only just begun (Kirfel, Icard & Gerstenberg, 2021). And this is where methodological pluralism comes in. We believe that using a diverse toolbox will get us closer to understanding what overall function causal judgments serve. Guided by the question of what purpose causal cognition serves, using a multitude of methods allows us to probe its function across a variety of causal behaviors such as predictions, explanations, judgments, and so on. Ultimately, such an approach is not restricted to causality, and should be extended to other forms of cognition. Recently, Lejarraga and Hertwig (2021) have suggested to combine both descriptive text protocols and experiential learning paradigms to get at a more comprehensive picture of human rationality: “The diverging conclusions about the mind drawn by proponents of the two research programs highlight the risk of relying exclusively on one class of experimental methods.”

Anything goes

“Anything goes!”, the philosopher Paul Feyerabend famously proclaimed about the use of scientific methods. In this blog post, we don’t want to go as far as to call for methodological anarchism. However, we do want to encourage exploring and exploiting the tools that empirical research hands us. Methodological pluralism is desirable specifically for theoretical reasons. In addition to studying how causal cognition works, we suggest investigating what purpose it serves. If we focus on causation in terms of its function, rather than what it is, the use of diverse methods might allow us to get closer at a unifying understanding of causation. This endeavour requires a concerted effort from philosophers and psychologists. In fact, philosophical theories of causation have played a decisive role for the development of psychological accounts of causal cognition, and there is already a similarly influential impetus from philosophy on the question of what causal cognition is for. Crucially, however, creating and evaluating evidence from diverse methodological paradigms will set the benchmark for however such a theoretical proposal might look like.

Lastly, we think that the argument for methodological pluralism is highly linked to and of relevance for the exchange between philosophy and the cognitive sciences more generally. Recently, a debate has sparked about the future and general research programme of philosophy. As Joshua Knobe has argued, “I would not say that contemporary philosophy is moving away from an attempt to go after the most fundamental questions. What is changing is people’s ideas about how to go after those questions.” And in a recent LSE philosophy blog post, Petr Jedlička suggests that an (even) closer connection to diverse empirical research might be one way for analytic philosophy to go forward. “[…] philosophy can also attempt to leave its safe haven of linguistic or textual analyses, or questionnaire surveys in the case of x-phi […] and embark on more daring voyages.”

There are many roads to causal cognition. We invite philosophers and psychologists alike to join us on this journey.

* * * * * * * * * * * *

Commentary

David Danks

* * * * * * * * * * *

Causation is widely recognized as an integral, perhaps even distinctively human, aspect of our cognitive lives. In many contexts, we cannot help but view the world in terms of causal relations, sometimes to our detriment (e.g., superstitions). Despite this centrality, though, psychological and philosophical theories and analyses of causal cognition are fragmented and divided, as Kirfel & Gerstenberg show in their valuable high-level overview of the current state of affairs.

Kirfel & Gerstenberg advance two main theses in response to this fragmentation. First, they argue that we psychologists and philosophers should use every available method to understand causal cognition. Researchers have increasingly used diverse methods in this field, but as they argue, we all can do better. Understanding causal cognition is a classic “hard problem,” and so we will undoubtedly need to use every available resource if we hope to make progress on it. At the same time, we will also need research on the different types of information provided by different methods, and the constraints that they imply on each other. The type of methodological pluralism advocated by Kirfel & Gerstenberg presupposes that we will eventually be able to combine the results—data, empirical phenomena, and so forth—revealed by the different methods. But such integration will require (substantial) additional research on our psychological and philosophical methods. If we want to avoid further fragmentation from “Anything goes!” in terms of our techniques and methods, then we need to study not only causal cognition, but also those methods themselves.

The second main thesis of Kirfel & Gerstenberg is likely more controversial: Research on causal cognition should include understanding the function of causal cognition, not just “what causation is (in the mind).” In general, functionalist accounts can highlight or reveal features that are otherwise neglected or hard-to-study. Moreover, they can provide substantial constraints on the cognitive processes and representations that (aim to) achieve the identified functions; there is no irreconcilable tension between functionalist and more mechanistic approaches to studying causal cognition. Kirfel & Gerstenberg argue specifically that we can make progress on understanding causal judgments about specific events by examining their function, particularly since it is not immediately obvious what that function might be.

An open question about this second thesis is whether the output of the functionalist approach should be understood as a means to develop other types of theories (e.g., by identifying constraints on mechanistic or rationalist models), or rather as an end in itself (e.g., a theory that need not be connected in any significant way with mechanistic accounts). Put somewhat differently, are Kirfel & Gerstenberg arguing that we should aim to develop functional analyses of causation and causal cognition, or provide functional definitions of them? The former would ideally connect with other theories about causal cognition, while the latter would not necessarily do so. If we adopt “Anything goes!” in terms of our approach, then are we giving up on the goal of a unified account of causal cognition?

One might also worry about the odds of success for a functional approach. In general, functionalist approaches work best when their target—causal cognition in this case—has only a small number of functions, ideally only one. If the target cognition could serve many different functions depending on context, goals, stimuli, and so forth, then our functionalist analysis can readily fragment into too many different pieces to be scientifically or philosophically useful. Kirfel & Gerstenberg consistently talk in the singular about the function or purpose, thereby implying that causal cognition (or at least, causal judgments about particular events) has a univocal function. But as they show, many different factors have been shown to impact causal judgments depending on the context, pragmatic needs, shared background knowledge, and much more. These empirical results suggest that causal judgments might serve multiple functions, depending on those contexts, needs, knowledge, and so forth. A functionalist approach to causal cognition might lead to insights, but we need to be careful not to focus on just one of the many plausible functions.

Kirfel & Gerstenberg persuasively argue for an “Anything goes!” expansion of the toolbox of methods, approaches, and theories that we use to study causal cognition. There has been substantial collective progress in this area, but we can (and should) do better as a community. At the same time, we should do so with our eyes open to the challenges of such an expansion. If we want to avoid the downside risk of further fragmentation, then we need to ensure that we have the knowledge and ability to integrate the results of different methods and approaches.

* * * * * * * * * * * *

References

Bechlivanidis, C., Schlottmann, A., & Lagnado, D. A. (2019). Causation without realism. Journal of Experimental Psychology: General, 148(5), 785.

Dinh, P. P. N., & Danks, D. (2021). Causal pluralism in philosophy: Empirical challenges and alternative proposals. Philosophy of Science, 88(5), 761-772.

Gerstenberg, T., Goodman, N. D., Lagnado, D. A., & Tenenbaum, J. B. (2021). A counterfactual simulation model of causal judgments for physical events. Psychological Review, 128(6), 936-975.

Griffiths, T. L., & Tenenbaum, J. B. (2005). Structure and strength in causal induction. Cognitive psychology, 51(4), 334-384.

Kirfel, L., Icard, T., & Gerstenberg, T. (2022). Inference from explanation. Journal of Experimental Psychology: General, 151(7), 1481–1501.

Lejarraga, T., & Hertwig, R. (2021). How experimental methods shaped views on human competence and rationality. Psychological Bulletin, 147(6), 535.

Lewis, D. (1973). Counterfactuals and comparative possibility. In Ifs (pp. 57-85). Springer, Dordrecht.

Leyssen, S. 2021. “Remaking ‘Michotte’: Reusing and Remaking Moving Images in the History of Perception Research.” Isis 112 (2): 315–25.

Lombrozo, T. (2010). Causal–explanatory pluralism: How intentions, functions, and mechanisms influence causal ascriptions. Cognitive psychology, 61(4), 303-332.

Michotte, A. (1946). The perception of causality, trans. R. Miles and E. Miles (London: Methuen & Co., Ltd., 1963).

Pearl, J. (2000). Causal inference without counterfactuals. Journal of the American Statistical Association, 95(450), 428-431.

Rottman, B. M., & Keil, F. C. (2012). Causal structure learning over time: Observations and interventions. Cognitive psychology, 64(1-2), 93-125.

Salmon, W. C. (1994). Causality without counterfactuals. Philosophy of Science, 61(2), 297-312.

Wolff, P. (2007). Representing causation. Journal of experimental psychology: General, 136(1), 82.

Woodward, J. (2007). Interventionist theories of causation in psychological perspective. Causal learning: Psychology, philosophy, and computation, 19-36.

Woodward, J. (2021). Causation with a human face: Normative theory and descriptive psychology. Oxford University Press.


Viewing all articles
Browse latest Browse all 26

Latest Images

Trending Articles





Latest Images