Baetu Tudor

One of the core tenets of scientific pluralism is that scientific knowledge will very likely remain fragmented into a multitude of partial models for the foreseeable future. This paper evaluates four lines of reasoning that support this claim by highlighting drivers of pluralism in the life sciences. More specifically, it is claimed that pluralism is likely to persist because all models must be partial as a matter of conceptual necessity, because the causal complexity of biological systems can only be accounted for by a plurality of partial models, because complex causal systems can be parsed into a multitude of incommensurable ways, and because the pragmatic success of pluralism supports the view that reality itself is plural. I argue that forecasts about the future state of scientific knowledge based on these considerations are biased, as they fail to factor in monistic trends mitigating pluralistic tendencies.

Multidisciplinary models aggregating ‘lower-level’ biological and ‘higher-level’ psychological and social determinants of a phenomenon raise a puzzle. How is the interaction between the physical, the psychological and the social conceptualized and explained? Using biopsychosocial models of pain as an illustration, I argue that these models are in fact level-neutral compilations of empirical findings about correlated and causally relevant factors, and as such they neither assume, nor entail a conceptual or ontological stratification into levels of description, explanation or reality. If inter-level causation is deemed problematic or if debates about the superiority of a particular level of description or explanation arise, these issues are fueled by considerations other than empirical findings.

An important strategy in the discovery of biological mechanisms involves the piecing together of experimental results from interventions. However, if mechanisms are investigated by means of ideal interventions, as defined by James Woodward and others, then the kind of information revealed is insufficient to discriminate between modular and non-modular causal contributions. Ideal interventions suffice for constructing webs of causal dependencies that can be used to make some predictions about experimental outcomes, but tell us little about how causally relevant factors are organized together and how they interact with each other in order to produce a phenomenon. I argue that lab research relies on more elaborated types of interventions targeting in a controlled fashion multiple variables at the same time in order to probe the temporal organization of causally-relevant factors along distinct causal pathways and to test for non-modular interaction effects, thus providing crucial spatial-temporal constraints guiding the formulation of more detailed mechanistic explanations.

A conflict between evidence for causation and the metaphysical requirement of spatiotemporal distinct causal relata arises if the results of experiments commonly described as ‘bottom-up’ and ‘top-down’ are taken to demonstrate ontological determination dependencies between parts and wholes or their respective behaviours. It has been argued that the problem can be circumvented if experimental results are interpreted in terms of relationships between variables measured and manipulated at separating scales. I argue that scale separation fails in the case of biological phenomena due to the small number of molecules and molecular events underlying these phenomena, which makes it impossible to completely disentangle micro and macro-scale behaviours. I defend instead a causal mediation interpretation according to which experiments assess the effects of interventions targeting the composition and exposure of biological systems. Under this interpretation, causal relata are ontologically distinct fractions or parts of biological systems, thus avoiding the oddity of part-whole causation.

It has been argued that intuition, perception and mechanistic rationales generate knowledge of singular causation unambiguously supporting the metaphysical view that causation is a local relationship between individual events. The analysis conducted in this paper contradicts this line of reasoning. Intuition, perception and mechanistic rationales rely, tacitly or explicitly, on prior contrasts between patterns describing populations of events. The fact that verdicts about singular causation are often reached in light of statistical, abductive or extrapolative inferences that are liable to error further entails that knowledge of singular causation is less certain than the knowledge of patterns and group-level difference-making dependencies from which it is derived. This indicates that claims about singular causation are not commonsense truisms, but complex epistemic constructs that should be approached more critically, especially if such claims are meant to guide action.

I argue that informational models of consciousness, including those proposed by the Integrated Information Theory, don’t presuppose or entail any particular view about the physical or metaphysical nature of consciousness. Such models only tell us how certain properties of consciousness can be mathematically described, thus providing a quantitative characterization of the phenomenon of consciousness that may contribute to the development of new methods of assessment and guide the explanatory project by supplying additional constraints on theoretical proposals. While informational models are in principle compatible with multiple physical interpretations, a preliminary argument in favour of the continuity between informational approaches and the mechanistic project in neuroscience can be made based on historical precedents.

David Chalmers advocates the view that the phenomenon of consciousness is fundamentally different from all other phenomena studied in the life sciences, positing a uniquely hard problem that precludes the possibility of a mechanistic explanation. In this paper, I evaluate three demarcation criteria for dividing phenomena into hard and easy problems: functional definability, the puzzle of the accompanying phenomenon, and the first-person data of subjective experience. I argue that none of the proposed criteria can accurately discriminate between the phenomenon of consciousness and mechanistically explainable phenomena.

It has been argued that supervenience generates unavoidable confounding problems for interventionist accounts of causation, to the point that we must choose between interventionism and supervenience. According to one solution, the dilemma can be defused by excluding non-causal determinants of an outcome as potential confounders. I argue that this solution undermines the methodological validity of causal tests. Moreover, we don’t have to choose between interventionism and supervenience in the first place. Some confounding problems are effectively circumvented by experimental designs routinely employed in science. The remaining confounding issues concern the physical interpretation of variables and cannot be solved by choosing between interventionism and supervenience.

The purpose of this article is to update and defend syntax-based gene concepts. I show how syntax-based concepts can and have been extended to accommodate complex cases of processing and gene expression regulation. In response to difficult cases and causal parity objections, I argue that a syntax-based approach fleshes out a deflationary concept defining genes as genomic sequences and organizational features of the genome contributing to a phenotype. These organizational features are an important part of accepted molecular explanations, provide the theoretical basis for a large number of experimental techniques and practical applications, and play a crucial role in in annotating the genome, deriving predictions and constructing bioinformatics models.

Causal mediation mechanisms are well supported by available experimental evidence and provide a practicable way to reductive physicalism. According to the causal mediation account of mechanistic explanation, descriptions as diverse as ‘black-box’ phenomena, mechanistic sketches and schemas mixing physically interpreted and operationalized biological, psychological and social variables, and detailed descriptions of mechanisms refer to the same causal structure circumscribed within the spatiotemporal boundaries of a replicable experimental setup. The coreference of coarser- and finer-grained descriptions of causal structures opens new possibilities for testing the reductive physicalism conjecture. I discuss experimental designs supporting the causal mediation account and show how recent studies demonstrating the biological mediation of mind-mind causal processes can provide evidence for reductive physicalism.

Causal pluralism can be defended not only in respect to causal concepts and methodological guidelines, but also at the finer-grained level of causal inference from a particular source of evidence for causation. An argument for this last variety of pluralism is made based on an analysis of causal inference from randomized experiments (RCTs). Here, the causal interpretation of a statistically significant association can be established via multiple paths of reasoning, each relying on different assumptions and providing distinct elements of information in favour of a causal interpretation.

Despite the consensus promoted by the evidence-based medicine framework, many authors continue to express doubts about the superiority of randomized controlled trials. This paper evaluates four objections targeting the legitimacy, feasibility, and extrapolation problems linked to the experimental practice of random allocation. I argue that random allocation is a methodologically sound and feasible practice contributing to the internal validity of controlled experiments dealing with heterogeneous populations. I emphasize, however, that random allocation is solely designed to ensure the validity of causal inferences at the level of groups. By itself, random allocation cannot enhance test precision, doesn’t contribute to external validity, and limits the applicability of causal claims to individuals.

The article analyses in detail the Meselson–Stahl experiment, identifying two novel difficulties for the crucial experiment account, namely, the fragility of the experimental results and the fact that the hypotheses under scrutiny were not mutually exclusive. The crucial experiment account is rejected in favour of an experimental-mechanistic account of the historical significance of the experiment, emphasizing that the experiment generated data about the biochemistry of DNA replication that is independent of the testing of the semi-conservative, conservative, and dispersive hypotheses. _1_ Introduction _2_ The Meselson–Stahl Experiment _3_ Some Difficulties for the Crucial Experiment Account _3.1_ Additional experiments were required _3.2_ Simplicity considerations are unconvincing _3.3_ The fragility of the experimental results _3.4_ The problematic interpretation of falsifyng results _3.5_ Not all possible hypotheses considered or tested _3.6_ The tested hypotheses were not mutually exclusive _4_ The Historical and Scientific Significance of the Meselson–Stahl Experiment _4.1_ The challenge for the crucial experiment account _4.2_ A mechanistic perspective on the experiment _4.3_ The experimental design and its independence from theoretical speculations _4.4_ The advantages of a ‘bottom-up’ mechanistic account _5_ Conclusion.

An important strategy in the discovery of biological mechanisms involves the piecing together of experimental results from interventions. However, if mechanisms are investigated by means of ideal interventions, as defined by James Woodward and others, then the kind of information revealed is insufficient to discriminate between modular and non-modular causal contributions. Ideal interventions suffice for constructing webs of causal dependencies that can be used to make some predictions about experimental outcomes, but tell us little about how causally relevant factors are organized together and how they interact with each other in order to produce a phenomenon. I argue that lab research relies on more elaborated types of interventions targeting in a controlled fashion multiple variables at the same time in order to probe the temporal organization of causally-relevant factors along distinct causal pathways and to test for non-modular interaction effects, thus providing crucial spatial-temporal constraints guiding the formulation of more detailed mechanistic explanations.

In this article, I argue that genomic programs are not substitutes for multi-causal molecular mechanistic explanations of inheritance, but abstract representations of the same sort as mechanism schemas already described in the philosophical literature. On this account, the program analogy is not reductionistic and does not ignore or underestimate the active contribution of epigenetic elements to phenotypes and development. Rather, genomic program representations specifically highlight the genomic determinants of inheritance and their organizational features at work in the wider context of the mechanisms of genome expression.

Different concepts define species at the pattern-level grouping of organisms into discrete clusters, the level of the processes operating within and between populations leading to the formation and maintenance of these clusters, or the level of the inner-organismic genetic and molecular mechanisms that contribute to species cohesion or promote speciation. I argue that, unlike single-level approaches, a multi-level framework takes into account the complex sequences of cause-effect reinforcements leading to the formation and maintenance of various patterns, and allows for revisions and refinements of pattern-based characterizations in light of the gradual elucidation of the causes and mechanisms contributing to pattern formation and maintenance.

This paper provides an account of the experimental conditions required for establishing whether correlating or causally relevant factors are constitutive components of a mechanism connecting input (start) and output (finish) conditions. I argue that two-variable experiments, where both the initial conditions and a component postulated by the mechanism are simultaneously manipulated on an independent basis, are usually required in order to differentiate between correlating or causally relevant factors and constitutively relevant ones. Based on a typical research project molecular biology, a flowchart model detailing typical stages in the formulation and testing of hypotheses about mechanistic components is also developed.

Emergent antireductionism in biological sciences states that even though all living cells and organisms are composed of molecules, molecular wholes are characterized by emergent properties that can only be understood from the perspective of cellular and organismal levels of composition. Thus, an emergence claim (molecular wholes are characterized by emergent properties) is thought to support a form of antireductionism (properties of higher-level molecular wholes can only be understood by taking into account concepts, theories and explanations dealing with higher-level entities). I argue that this argument is flawed: even if molecular wholes are characterized by emergent properties and even if many successful explanations in biology are not molecular, there is no entailment between the two claims.

Multidisciplinary models aggregating ‘lower-level’ biological and ‘higher-level’ psychological and social determinants of a phenomenon raise a puzzle. How is the interaction between the physical, the psychological and the social conceptualized and explained? Using biopsychosocial models of pain as an illustration, I argue that these models are in fact level-neutral compilations of empirical findings about correlated and causally relevant factors, and as such they neither assume, nor entail a conceptual or ontological stratification into levels of description, explanation or reality. If inter-level causation is deemed problematic or if debates about the superiority of a particular level of description or explanation arise, these issues are fueled by considerations other than empirical findings.

The new mechanistic philosophy is divided into two largely disconnected projects. One deals with a metaphysical inquiry into how mechanisms relate to issues such as causation, capacities and levels of organization, while the other deals with epistemic issues related to the discovery of mechanisms and the intelligibility of mechanistic representations. Tudor Baetu explores and explains these projects, and shows how the gap between them can be bridged. His proposed account is compatible both with the assumptions and practices of experimental design in biological research, and with scientifically accepted interpretations of experimental results.

Current debates surrounding the virtues and shortcomings of randomization are symptomatic of a lack of appreciation of the fact that causation can be inferred by two distinct inference methods, each requiring its own, specific experimental design. There is a non-statistical type of inference associated with controlled experiments in basic biomedical research; and a statistical variety associated with randomized controlled trials in clinical research. I argue that the main difference between the two hinges on the satisfaction of the comparability requirement, which is in turn dictated by the nature of the objects of study, namely homogeneous or heterogeneous populations of biological systems. Among other things, this entails that the objection according to which randomized experiments fail to provide better evidence for causation because randomization cannot guarantee comparability is mistaken.

The paper discusses methodological guidelines for evaluating mechanistic explanations. According to current accounts, a satisfactory mechanistic explanation should include all of the relevant features of the mechanism, its component entities and activities, and their properties and organization, as well as exhibit productive continuity. It is not specified, however, how this kind of mechanistic completeness can be demonstrated. I argue that parameter sufficiency inferences based on mathematical model simulations provide a way of determining whether a mechanism capable of producing the phenomenon of interest can be constructed from mechanistic components organized, acting, and having the properties described in the mechanistic explanation.

Both clinical research and basic science rely on the epistemic practice of extrapolation from surrogate models, to the point that explanatory accounts presented in review papers and biology textbooks are in fact composite pictures reconstituted from data gathered in a variety of distinct experimental setups. This raises two new challenges to previously proposed mechanistic-similarity solutions to the problem of extrapolation: one pertaining to the absence of mechanistic knowledge in the early stages of research and the second to the large number of extrapolations underpinning explanatory accounts. An analysis of the strategies deployed in experimental research supports the conclusion that while results from validated surrogate models are treated as a legitimate line of evidence supporting claims about target systems, the overall structure of research projects also demonstrates that extrapolative inferences are not considered definitive or sufficient evidence, but only partially justified hypotheses subjected to further testing. 1 Introduction2 Surrogate Models2.1 What exactly is a surrogate model?2.2 Why use surrogate models?3 Prior Validation of Surrogate Models3.1 The validation and ranking of surrogate models in the early stages of basic research3.2 The validation of surrogate models in later stages of basic research and in clinical research4 ‘Big Picture’ Accounts and the Extrapolations Underpinning Them4.1 The mosaic nature of mechanistic descriptions in basic science4.2 Challenges for mechanistic-similarity-based validation protocols5 Retrospective Testing of Extrapolated Knowledge5.1 Holistic confirmation5.2 Fallback strategies6 Conclusions.