Complex networks, structural explanations, and the role of values in experimental linguistics

PRINCIPIA 28(4): 517–540 (2024) doi: 10.5007/1808-1711.2024.e92531 Published by NEL — Epistemology and Logic Research Group,Federal Universityof Santa Catarina(UFSC),Brazil. COMPLEX NETWORKS, STRUCTURAL EXPLANATIONS,AND THE ROLE OF VALUES IN EXPERIMENTAL LINGUISTICS JUAN BAUTISTA BENGOETXEA Universidad del País Vasco/Euskal Herriko Unibertsitatea, SPAIN [email protected] https://orcid.org/0000-0003-1158-1122 Abstract. First, a brief state of the art on the role of values in science is presented. The main points of the critique of the ideal of a free-value science are depicted (Sec. 2), followed by a framework of four phases in the development of scientific activity that allows us to better assess the influence of values in science. The basic thesis of the present text criticizes the conception close to Heather Douglas’ thesis, according to which values, including cognitive values, can only play an indirect role in science (Sec. 3). In order to show that, at certain phases of science, the explanatory power (a cognitive value) can play a direct role, we propose an examination of structural explanation, as distinguished from mechanistic explanation, in a case of experimental linguistics related to language acquisition (Sec. 4). The conclusion is presented in the last section. Keywords: complex networks •structural explanation •experimental linguistics •cognitive values RECEIVED: 18/01/2023 REVISED: 08/12/2023 ACCEPTED: 22/05/2024 1. Introduction The social relevance of philosophy continues to be a topic of debate in our discipline. According to Kitcher (2023, p.27), the philosophy of science and the epistemology of science would be quite well placed on this landscape, which unfortunately shows a shift away from public interest in philosophical products. Haack (2016) argues that interest could be revived precisely if we were to place philosophy in contexts of more social relevance. In this respect, she agrees with the results projected by authors such as Douglas (2009b, 2014), Lacey (2004, 2011), Martins Dos Reis and De Patta Pillar (2018), to mention a few cases of a philosophy directly linked to issues whose interest goes beyond the boundaries of the academy. Obtaining and managing reliable knowledge is an inherent human goal and its social relevance is actually very important. If we consider that the highest form of doing this is the scientific one, its products, generally speaking (theories, hypotheses, models, technologies), will have to be reliable to be used in our task of understanding © 2024 The author(s). Open access under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License. 518 Juan Bautista Bengoetxea phenomena, explaining and predicting facts and hypotheses, intervening and influencing research domains with a minimum guarantee, or controlling various variables in these domains (experimentation) (see Hacking 1983, p.149–52; Douglas 2009b, p.93–4, 100–2, 112–3; Cartwright et al. 2022). In the case we are concerned with here, scientific explanation, its targets (explananda) depend to a large extent on all these products of science and also on the values underlying the scientific enterprise in situations—all or almost all—in which uncertainty and risk are part and parcel of the work that must be done (Douglas 2014). In these very common circumstances, researchers, regulators and science policy makers have to choose certain research pathways from a wider range of options. It is therefore very usual to wonder about the criteria, the interest or the objective that makes scientific research focus on one object or another, on one strategy or another, with such social weight. Any response will undoubtedly take the form of a value judgement or depend on value judgements.1The ideal of a value-free science had a major impact on the philosophy of science and the epistemology of science for decades (see Rezaee & BikaraanBehesht 2023). However, that ideal was put into crisis and a major shift towards the examination of non-epistemic and non-cognitive values was undertaken (see Lacey 1999, 2004; Douglas 2009b). This change led to the almost consensual endorsement of the belief that ethical and social values (these would be the non-epistemic/noncognitive ones) do influence scientific activity. Current issues as important for science and society as regulation and decision making, in addition to drawing on scientific work based on data and evidence, must take into account the influence of values and interests of the stakeholders involved.2 The shift away from the ideal of a value-free science has, however, led us to radically opposed positions, to the extent that the role of epistemic and cognitive values as such has been largely downplayed. It seems that the shift towards the study of ethical and social values has been excessively skewed towards certain biases and even positions that deny or greatly reduce the role of epistemic and cognitive values (Harding 1986, Longino 1990, see Haack 1998). This could call into question the very nature of the scientific enterprise. From a moderately pro-epistemic and procognitive approach (see Risjord 2023), we consider it important to partially return to a defence of the positive role of epistemic and cognitive values, especially in a human science such as linguistics (Sec. 4). To ponder the plausibility of the thesis that ethical and social values dominate in the social and human sciences, we first look at an epistemological proposal, specific to the philosophy of science, which projects an image of science as constituted in four phases (Lacey 1999, Martins Dos Reis & De Patta Pillar 2018, Douglas 2009b).3 It is within these phases that we can inquire into the role of epistemic, cognitive and social values, and into which, if any, are strongest in each phase. The interest of this lies in the fact that we are interested in assessing scientific activity in a social and PRINCIPIA 28(4): 517–540 (2024) Networks, Explanations and Values in Linguistics 519 human science such as linguistics, a field in which, according to the advocates of the role of social and ethical values, these acquire even greater importance, if possible. However, here we want to argue that it is precisely epistemic and cognitive values, in particular the explanatory capacity, that play a crucial role in this science, and that they dispel some of misunderstandings surrounding the multiplicity of interpretations and the weakness of the evidence used. Building on Douglas’s (2009a, 2009b) dissection of the (direct and indirect) role of values in science, our analysis of the case of structural explanation in linguistics seeks to nuance that response and to re-emphasize the direct role of some cognitive values in scientific activity, especially that of explanatory power (see Cartwright et al. 2022). 2. A science with social value Despite the aforementioned shift towards the study of non-epistemic and non-cognitive values, it is still possible to argue that there are epistemic and cognitive values that influence scientific activity in a very relevant way. Heather Douglas (2009b, p.92) starts from an interesting typology of values. According to her, we can distinguish between ethical values and social values on the one hand, and epistemic values and cognitive values, on the other. The latter two types differ in that epistemic values as such are not really values, but criteria that any scientific enterprise must meet. These are established because science is valued as such. They therefore fix the scope of science from the outside, and do not act as values within science. Cognitive values, on the other hand, constitute something more precise than a loose grouping of acceptable values in science (this is what is often equated with epistemic values). Cognitive values, he points out, are those elements of scientific work that helps us to reflect on the evidential and inferential features of theories and data that scientists hold (Douglas 2009b, p.93). They would be assistants to the cognitive work of scientists. In short, if cognitive values are concerned with the fruitfulness of research, epistemic values are concerned with the starting or basic goal of research, namely the true or, more usually, reliable knowledge. Furthermore, and importantly, cognitive, social and ethical values should never claim to play a direct role in science. That is, they should not be equated with the evidence and data used by scientists. We see, therefore, that a value of high epistemic incidence is the reliability of the products of science. Cartwright et al. (2022, Part I, 3), in the wake of an idea of Otto Neurath, point out that reliability is an umbrella concept or Ballung. The concepts linked to cognitive values (scope, simplicity, explanatory capacity, consistency) would be deployed alongside it.4Science, among other things, would seek to propose coherent models, hypotheses or theories, or to offer adequate or useful theories. But we also know that theories and the scientific practices associated with them have PRINCIPIA 28(4): 517–540 (2024) 520 Juan Bautista Bengoetxea degrees of uncertainty and are fallible and provisional (see Shrader-Frechette 1993), a fallibility that scientific practices and theories seek to reduce. What does seem to be accepted is that there is no such thing as a value-free science (whether or not that ideal is upheld), on the one hand, and that, although it is value-laden, science still has its own virtuous features (reliability, truth) without which science would cease to be science and research would cease to be research. In addition to those mentioned by Douglas (see endnote 4), we find precision, clarity, justification, objectivity, truth, empirical adequacy (see van Fraassen 1980; Lacey 1997, p.7), respect for evidence, as well as the effort to minimize fallibility and uncertainty (Douglas 2009b, p.95). Ethical and social values also influence science. Ethical values are concerned with what is good or right, and are very important in examining the consequences of scientific errors for the general public. Among other things, they help us to weigh up whether potential benefits are actually potential dangers to be taken into account or not, whether there are dangers to be assumed without any price, etc. (Douglas 2009b, p.92). Social values arise from what a particular society values: justice, privacy, freedom, social stability, innovation, etc. (Lacey 1999, p.28–9). Taking these different types of values seriously has promoted the aforementioned critique of the ideal of a value-free science (Kincaid; Dupré; Wylie 2007), which has changed the overall scope of the debate. Both the supremacy of the old value-free ideal and the exclusive interest in epistemic values (McMullin 1982; Laudan 1984) of scientific research have been called into question (Douglas 2009b, p.89). At the present it is social, political or ethical values that are of considerable interest in recent axiological debates (Koskinen & Rolin 2022; McMullin 1982). The efforts to overcome the ideal of a science free of values has been made in order to adopt an increasingly pluralistic scientific attitude, both in natural and social sciences (Kellert; Longino; Waters 2006). This new attitude is based on the basic idea that scientific research is a human activity that is goal-oriented and highly influenced by nonepistemic values and norms of research settings (contexts) (González 2013, p.1505). The four most prominent lines of criticism of the ideal of a value-free science can be summarized as follows: [1]Philosophers sought to ‘unmask’ the so-called epistemic values that are in fact non-epistemic ones. Longino (1996) has argued that so-called ‘epistemic’ values are not in fact exclusively epistemic, for their use assumes political and social values in contexts of scientific judgement. There are values that underlie scientific judgement and they are not always, or even usually, politically neutral. Longino proposes to juxtapose these values with values (including many of a feminist nature) such as novelty, ontological heterogeneity, mutuality of interaction, applicability to human needs and diffusion of power. Failure to recognise this axiological richness and variety can lead to bias and adverse research outcomes (Wilholt 2009, p.96). PRINCIPIA 28(4): 517–540 (2024) Networks, Explanations and Values in Linguistics 521 [2]Richard Rudner’s (1953) now classic critique (see Resnik & Elliott 2023, Lacey 1999, p.71–4) questioned the idea that any decision about a scientific theory or hypothesis can be value-free (Douglas 2009b, p.51-2). According to him, any acceptance or rejection of a hypothesis is set against a background of knowledge and priorities (Rudner 1953, p.2), and a scientific hypothesis is never confirmed beyond reasonable doubt: there is always the possibility of a decision being wrong (Resnik & Elliott 2023, p.7). The decision to accept or reject a hypothesis implies a value judgement (at least implicitly); one has to judge which of the consequences of an erroneous decision is more acceptable.5Hence, non-epistemic (moral, social) values are a necessary part of the scientist’s core activity of accepting and rejecting hypotheses. [3]Moreover, using a strategy similar to Longino’s (1996), Putnam (2002, p.140– 2) rejects the possibility of a value-free science by means of a semantic argument that attacks the neutrality of scientific theories. Putnam (2002) points out that among non-epistemic and non-cognitive concepts, especially among ethical ones, there are values that can be considered both normative and descriptive depending on the perspective adopted. These are cases like ‘cruel’, ‘dirty’ or the like. He calls them ‘thick’ concepts and uses them precisely to provide counterexamples to the fact/value distinction or, in our context, non-epistemic value/epistemic value.6If the use of terms such as these intertwining facts and values is unavoidable in scientific practice and reasoning, any proposal of scientific hypotheses and results can hardly be value-free. This seems to undermine any thesis advocating the impartiality of science and the neutrality of its values.7 [4]Some philosophers of science argue that non-epistemic values have an indispensable legitimate role to play in assessing the risks involved in the acceptance of scientific hypotheses and decisions (the argument from inductive risk) (Elliott & Richards 2017; Elliott & Steel 2017, Part III; Brown 2013; Douglas 2000, 2009b, p.58; Rolin 2015, p.161; Steel 2010; Wilholt 2009; see Rudner 1953, p.2). One of its premises is that the acceptance or rejection of hypotheses (the usual task of scientists) involves uncertainty. In doing so, a scientist has to decide whether the available evidence is strong enough to warrant acceptance. This decision depends on the risks involved. If scientists accept a false hypothesis, there may be a cost associated with this type of error. If they reject a true hypothesis, the cost will be of a different kind. The key premise of this critique is that the assessment of the costs involved in both types of errors is a matter of moral value judgement (Rudner 1953, p.3). As can be seen, it is assumed that defending the role of non-epistemic and noncognitive values in attempting scientific hypotheses or theories does not necessarily rule out the possibility of creating regulatory frameworks in which beneficial and harmful consequences of such values could be distinguished (Crespo 2019; DiekPRINCIPIA 28(4): 517–540 (2024) 522 Juan Bautista Bengoetxea mann & Peterson 2013). This makes it possible to propose alternatives, sometimes quite complex, to the ideal according to which science is free of values. Those would be alternatives that should redefine and distinctly locate several epistemic notions such as reliability, objectivity, and rationality (Douglas 2009b, Chapter 6; Koskinen 2021; Catwright 2022).8These epistemic values guide the task performed together with cognitive values, including the explanatory power (Cartwright et al. 2022; Pritchard 2021, p.5523; Intemann 2015; see Psillos 2015; Koskinen & Rolin 2022, p.193). The combined form of all the products derived from taking these values into consideration would be that of a network. The location of values, in turn, would not be given by the classical distinction between epistemic and cognitive versus social and ethical values, but rather by the role played by each type or each token: direct or indirect (Douglas 2009b, p.95–7). But we point out the relevance of the ‘token’ precisely because, contrary to what Douglas claims, we consider that the role of the explanatory power of scientific theories (understood as a cognitive value) cannot be labelled as ‘indirect’ tout court. Both mechanistic explanations and (mathematic-like) structural explanations in experimental linguistics will serve as an illustrative case for our critique of Douglas. In order to locate and seek to understand this critique, we must see that the acceptability of scientific activity would therefore be based on its reliability, related to the use of evidence, together with the assessment of the lack of bias, the significance of scientific theories and all their products, their completeness, their consistency and other cognitive values. The constant consideration of these epistemic and cognitive values in scientific activity is not exempt from the participation of social and ethical values as well. The network in which they play a part is complex and tangled, and it is therefore useful to unravel, at least partially, those phases of the scientific enterprise where the use of values is more or less direct. The complex nature of scientific activity should be emphasised if we were to understand its products in any of the phases in which it takes place. Understanding this kind of complexity is linked to the reliability we expect from science from the very moment we begin to develop it (Cartwright et al. 2022; Brigandt 2015; Potter 2006; Lacey 1999, p.38–9). 3. Values into scientific research fundamental phases Values are inherent in scientific practice aimed at providing reliable knowledge (McIntyre 2019, p.230; Douglas 2009b, p.113–4; Potter 2006, p.78).9How to articulate such reliability rolled in values, including social and ethical ones, is a pressing goal for the philosophy of science (see Cartwright et al. 2022, Part II, 4). If values, therefore, cannot be refused from scientific activity, then reliability or, as it may be, approximation to truth, would have to be obtained through different strategies in PRINCIPIA 28(4): 517–540 (2024) Networks, Explanations and Values in Linguistics 523 scientific research.10 One consists of recognizing the plurality of values at the different stages of research, from the very beginning, when researchers select their goals and working methods—theory-guessing, making conjectures or hypotheses, and so on (Vilhalemm 2016; Popper 1994, p.95))—, to the final implementation stage. In all these stages, the weight of values can be estimated in some way, and their linkage to properly epistemic strategies based on evidence and data can be calibrated. It is interesting to critically examine the roles that values can play in scientific activities (Lacey 2011; Lacey & Mariconda 2014).11 This examination starts from the understanding that scientific activity can be structured in phases, specifically in four (’Ph’ onwards) (see Douglas 2009b, p.88–9): [Ph-1]Adoption and development of a research strategy; [Ph-2]gathering data; [Ph-3]cognitive and/or epistemic evaluation of the theories, models or hypotheses proposed to solve some problem identified in the initial adoption; and [Ph-4]implementation of the obtained scientific knowledge. [Ph-1]Adoption and development of a research strategy: The first decision concerns the research strategy; the research topic must be decided. One usual starting point of a research is to propose some conjecture, hypothesis, problem or theory (Cartwright 2022, p.19–20). That point serves to select and shape the phenomena that scientists intend to investigate. In order to do this, researchers irremediably take non-cognitive values into account. Adopting a research strategy is similar to propose a complex framework of scientific inquiry made up of various elements, among which the identification of an initial problem and of a working guide hypothesis helping to solve the former are two remarkable ones (Lacey & Mariconda 2014, p.647). This initial guide to scientific inquiry is crucial in order to establish the relevant types of empirical data, the appropriate descriptive categories for making observational reports, as well as the theories linked to data. The link between data and theories, it must be emphasized, requires the adoption of a strategy from the beginning. [Ph-2]Gathering data: Once the object of research has been determined, scientists enter the justification phase, in which data and evidence are collected and accumulated,12 attempts are made to justify certain hypotheses, to explain facts in the field studied or, in some cases, to predict those facts. The ways of collecting evidence can be very varied if we compare the natural sciences with the social sciences. In the latter, it is very common to appeal to interviews, reports, field work and, in general, to tools that are foreign to most of the natural sciences (Douglas 2009b, p.88). [Ph-3]Epistemic evaluation of theories (models, hypotheses): Once the necessary efforts have been made to confirm and verify (or disprove or falsify) certain hypotheses, the strongest or the best supported hypothesis is chosen. To confirm a hypothesis is to find evidence or reasons for it and not against it, so that the hypothesis likelihood increases. However, at this stage, contrary to what Douglas (2009b) advocates, cognitive values can act as criteria for epistemic evaluation and be independent from PRINCIPIA 28(4): 517–540 (2024) 524 Juan Bautista Bengoetxea non-cognitive values (McIntyre 2019, Landes 2020). As theories or hypotheses tend to be accepted or refuted more clearly according to some criterion of reliability, a theory will be accepted for a field of phenomena if it exhibits its own cognitive values (scope, accuracy, explanatory power, predictive power, etc.) not shown by any alternative theory—or shown by the latter in a lesser degree. This is always made in the light of empirical data and the relevant and sufficient evidence collected from observation, experimentation or even simulations, and always on the basis of the used theory and of the phenomena to be studied. Non-cognitive values and certain non-empirical (metaphysical, speculative) convictions play no role in the ‘Ph-3’ of the theory assessment choice. This kind of values hardly, or not at all, determines the investigation in Ph-3 (see Betz 2013, p.212). [Ph-4]The implementation of knowledge: To the extent that a hypothesis or theory is consolidated in a scientific discipline, and depending on researchers’ interests and values, the conclusions and the practical, applicable and even regulatory options are analyzed and selected for decision-making in science policy (see Bengoetxea & Todt 2021, p.57–62; Bengoetxea 2024, p.113). This is also called ‘the phase of implementation’ of science (see González 2013). It may adopt the form of a transition from the natural sciences’ context to the context of social sciences (from biochemistry to regulation, typically). By applying scientific knowledge, some ideals are projected and researchers serve certain interests that reflect specific non-cognitive values. Science is applied because scientists both expect some benefits stimulated by their initial research interests and assume to overcome some potential negative consequences of scientific applications (Lacey & Mariconda 2014). It is not a necessary condition that these four phases develop in a successive order (for example, Ph-2 and Ph-3 could influence Ph-1, or the processes in Ph-4 could (allegedly) affect previous phases). Since the processes of obtaining scientific knowledge and its application are subject to varying degrees of uncertainty, science is assumed to be fallible (this is also a condition for doing science today). It means that scientists sometimes have to make decisions (or hypotheses) under risky conditions (Elliott & Richards 2017, p.2). In doing so, it is most usual to think about the (also social) interests of the cognitive enterprise, as well as the possible errors of such decisions. Thinking about consequences requires taking into account social, political, ethical and economic values, all of which can influence the process of verification and justification of a hypothesis by the scientific community. Therefore, it seems to us that the Ph-1/Ph-4 scheme is a good tool that could make easier the representation of the cognitive processes established and developed in science. The case study of structural explanation (section 4) is proposed to support this hypothesis. In the case of experimental linguistics, structural explanatory techniques involve introducing a cognitive value (explanatory power) in Ph-2 and Ph-3 as a fundamental element. This does not preclude the need to resort to non-cognitive values in some cases. In this PRINCIPIA 28(4): 517–540 (2024) Networks, Explanations and Values in Linguistics 525 sense, as we approach the social or human sciences, where human behavior is the object of research, it is more difficult to avoid the influence of non-cognitive values. Human language is a very complex phenomenon with many edges. Therefore, the articulation and evaluation of evidence and hypotheses turns out to be very complex and allows for different interpretations that may be motivated by multiple values or interests, at least in principle, although the main goal is their reliability as an epistemic value that conditions scientific activity. But explanatory power—conceived as a cognitive value playing a direct role in Ph-2 and Ph-3—is located in a complex network along with other values—cognitive, ethical, and social—and can help develop a more reliable and more integral science (Douglas 2009b, p.95), also in the case of experimental linguistics. 4. Complex networks, structural explanations and the study of some features of language Hempel’s nomological-deductive model of explanation ceased to be the quintessential approach in disciplines such as biology and neurosciences studying living and cognitive systems (Kitcher 1989; Douglas 2010; Hempel 1965; see Diez; Khalifa; Leuridan 2013; Douglas 2009a, p.449; Salmon 1989, p.94–101; Friedman 1974). Instead of seeking universally applicable laws or predictive models, researchers into these fields have sought to understand the behavior of living organisms by analyzing them in parts (structures). This explanation strategy was based on the study of mechanisms (Machamer; Darden; Craver 2000; Moreno & Suárez 2020, p.146). However, mechanistic explanation is a type of explanation that cannot always account for certain aspects of the biology of organisms (development, reproduction, etc.) (Alleva; Díez; Federico 2017) nor deal with other properties of complex systems that are better explained by tools from complexity sciences (Huneman 2018, Brigandt; Green; O’Malley 2017). Something can be said to be complex if it is made up of interconnected parts and if the understanding of its behavior requires comprehending not only the behavior of its parts, but also the way in which they mutually act to give rise to an overall behavior. Global description requires describing the parts, and these in turn are described in terms of the whole (Bar-Yam 1997, p.1). The most important characteristics to be taken into account of complex systems can be summarized as follows (Bar-Yam 1997, p.5): (1) the elements or entities of a system (and their quantity), (2) the interactions among them (and their strength), (3) their formation and function, (4) their diversity, (5) their environment, and (6) their activities. But what is complexity? To detail a description of a complex system, it is useful to understand how the concept of complexity is related to that of emergence (Bar-Yam 1997, p.6). PRINCIPIA 28(4): 517–540 (2024) 532 Juan Bautista Bengoetxea direct role in accepting or rejecting a hypothesis (Douglas 2009b, p.110). She claims that neither cognitive nor social values are a good reason by themselves to accept or not accept some theory. We partially disagree with this statement. The explanatory power of a linguistic theory is one element among others (always combined with other elements (see Cartwright et al. 2022)) that sometimes plays only an indirect role (guiding and supporting decision-making by researchers or regulatory scientists),18 but at other times can act directly because scientific explanatory power, unlike other cognitive values, is based on the development of explanations themselves, which is directly linked to working evidence and data. This normally occurs in phases Ph-2 and Ph-3, although it is related to the other phases. What we do not demand here is that the explanation has to be based on mechanisms, in general, nor on causal mechanisms, in particular. A structural explanation, based on mathematical topology, can provide a representation of an explanandum that reflects a network and that can be added to help solve several problems. 5. Conclusion The openness of philosophy to other fields of interest takes the form of a focus on issues of both social and scientific interest. To this end, it is essential to work hand in hand with scientists themselves (Douglas 2010, p.317). Some philosophers have claimed that philosophy of science should regularly get engaged into scientific practices in order to understand their underlying philosophical issues (see Kitcher 2023, p.16). It is about a kind of social relevance that we might describe as ’being philosophically attentive’ to those aspects of science that are particularly important to society. Given the desideratum that philosophy should obtain greater relevance among scientists and the general public (Kitcher 2023, p.16), we have considered one nexus connecting just the three domains of values. The values—whether epistemic, cognitive, social, or ethical—were initially conceived (according to the ideal of a value-free science) as an undesirable influence on the scientific enterprise and on its integrity, objectivity, and reliability. However, these concepts themselves already took the form of values, at least epistemic and cognitive ones. A space was made for them, but not for the supposedly harmful ‘social and ethical values’. This image was soon superseded in the philosophy of science by a series of well-established critical arguments, originated by Rudner in 1953, so that more detailed and complex approaches, such as that of Heather Douglas or Hugh Lacey, among others, established phases of science in which the integration of values gradually varied. Thanks to this new sharing of the complex procedures of making theories and practices in science, it has been possible to understand that the role of values is located in some combined processes PRINCIPIA 28(4): 517–540 (2024) Networks, Explanations and Values in Linguistics 533 similar to what Neurath meant with his ’Ballung concepts’. In these circumstances, Douglas further incorporated the interesting idea that values, of whatever kind, may influence various phases of science, but will never play a direct role (Douglas 2010, p.328; 2009b, p.96, 112). Well, this is precisely the thesis that we criticize here. The cognitive value ‘explanatory power’ is essential to support the epistemic value ‘reliability’. And scientific explanation is not limited to playing an indirect guiding role in decision making, but, in the sense of the Ballung mentioned above, it is developed together with the handling of data, evidence and other properly scientific products and means. We have tried to show this in the case of experimental linguistics. We have proposed the notion of structural explanation, based on mathematics, which serves as a complement to many results of mechanistic explanations, since structural explanations are not conceived as mere ‘aids’ for decision making in the central phases of science (Ph-2 and Ph-3, for example). Later we have moved towards those three desiderata through the application of a structural explanatory approach. The task of elaborating a notion of scientific explanation is properly philosophical and, in order to carry it out, we have also taken into account an approach based on a direct role of cognitive values. We have presented a case from experimental linguistics with which we work in close proximity to values and evidences linked to language acquisition. We think that this is a sufficient demonstration, at the moment, to positively value the scientific (linguistic) and social advances of an experimental science in continuous progress and helped by a philosophy of science, epistemologically rooted, that attempts to adequately conceptualize it. References Abbuhl, R.; Gass, S.; Mackey, A. 2013. Experimental research design. In R. Podesva; D. Sharma (ed.), Research Methods in Linguistics, p.116–134. Cambridge: Cambridge University Press. Alleva, K.; Díez, J.; Federico, L. 2017. Models, theory structure and mechanisms in biochemistry: The case of allosterism. Studies in History and Philosophy of Biology and Biomedical Sciences 63: 1–14. Antony, L. 1993. Quine as feminist: The radical import of naturalized epistemology. In L.M. Antony; C.E. Witt (ed.), A Mind of One’s Own: Feminist Essays on Reason and Objectivity, p.110–153. Boulder: Westview. Bar-Yam, Y. 1997. Dynamics of Complex Systems. Reading, MA.: Addison-Wesley. Barabási, A.-L. 2002. Linked: The New Science of Networks. Cambridge, MA: Perseus. Barceló-Coblijn, L.; Corominas-Murtra, B.; Gomila, A. 2012. Syntactic trees and small-world networks: syntactic development as a dynamical process. Adaptive Behavior 20(6): 427– 442. PRINCIPIA 28(4): 517–540 (2024) 534 Juan Bautista Bengoetxea Barceló-Coblijn, L.; Duguine, M.; Irurtzun, A. 2019. The Emergence of Hubs in Complex Syntactic Networks and the DP Hypothesis: The Relevance of a Linguistic Analysis. In À. Massip; G. Bel-Enguix; A. Bastardas-Boada (ed.), Complexity Applications in Language and Communication Sciences, p.273–288. London: Springer. Barceló-Coblijn, L.; Serna, D.; Isaza, G.; Castillo, L.F.; Bedia, M. 2017. Netlang: A software for the linguistic analysis of corpora by means of complex networks. PlosOne 12(8): 1–15. Barker, G., Kitcher, P. 2014. Philosophy of Science: A New Introduction. Oxford: Oxford University Press. Bechtel, W.; Richardson, R. C. 1993. Discovering complexity: Decomposition and localization as strategies in scientific research. Princeton: Princeton University Press. Bengoetxea, J. B. 2024. Imparcialidad y demarcación de valores en la actividad científica. Revista Iberoamericana de Ciencia y Tecnología 19(55): 107–125. Bengoetxea, J. B.; Todt, O. 2021. Decision-Making in the Nutrition Sciences: A Critical Analysis of Scientific Evidence for Assessing Health Claims. Manuscrito 44(3): 42–69. Betz, G. 2013. In defence of the value free ideal. European Journal for Philosophy of Science 3: 207–220. Brigandt, I. 2015. Social values influence the adequacy conditions of scientific theories: beyond inductive risk. Canadian Journal of Philosophy 45: 326–356. https://doi.org/10.1080/00455091.2015.1079004. Brigandt, J.; Green, S.; O’Malley, M. A. 2017. Systems biology and mechanistic explanation. In S. Glennan; P. Illari (ed.), The Routledge handbook of mechanisms and mechanical philosophy, p.362–374. London: Routledge. Brown, M. J. 2013. Values in Science beyond Underdetermination and Inductive Risk. Philosophy of Science 80(5): 829–839. Bueno, O. 2011. Partial Truth and Visual Evidence. Principia 15(2): 249–270. Cartwright, N. 2022. A Philosopher Looks at Science. Cambridge: Cambridge University Press. Cartwright, N.; Hardie, J. 2012. Evidence-based policy: a practical guide to doing it better. New York: Oxford University Press. Cartwright, N.; Hardie, J.; Montuschi, E.; Soleiman, M.; Thresher, A.C. (ed.). 2022. The Tangle of Science: Reliability Beyond Method, Rigour, and Objectivity. Oxford: Oxford University Press. Chomsky, N.; Miller, G.A. 1963. Introduction to the formal analysis of natural languages. In L. Duncan; R.R. Bush; E. Galanter (ed.), Handbook of mathematical psychology, p.269–321. New York: John Wiley. Corominas-Murtra, B. 2007. Network statistics on early English Syntax: Structural criteria. arXiv e-print.https://arxiv.org/abs/0704.3708. Access 12/12/2022. Corominas-Murtra, B.; Valverde, S.; Solé, R.V. 2009. The ontogeny of scale-free syntax networks: Phase transitions in early language acquisition. Advances in Complex Systems (ACS) 12: 371–392. Coyte, K. Z.; Schluter, J.; Foster, K. R. 2015. The ecology of the microbiome: Networks, competition, and stability. Science 350(6261): 663–666. Crespo, R.F. 2019. Liberal Naturalism and Non-epistemic Values. Foundations of Science 24: 247–273. Deulofeu, R.; Suárez, J.; Pérez-Cervera, A. 2021. Explaining the behaviour of random ecological networks: The stability of the microbiome as a case of integrative pluralism. Synthese PRINCIPIA 28(4): 517–540 (2024) Networks, Explanations and Values in Linguistics 535 198(3): 2003–2025. Diekmann, S.; Peterson, M. 2013. The Role of Non-Epistemic Values in Engineering Models. Sci. Eng. Ethics 19: 207–218. Díez, J. A.; Khalifa, K.; Leuridan, B. 2013. General theories of explanation: Buyers beware. Synthese 190: 379–396. Douglas, H. E. 2000. Inductive Risk and Values in Science. Philosophy of Science 67(4): 559– 579. Douglas, H. E. 2009a. Reintroducing Prediction in Explanation. Philosophy of Science 79: 444–463. Douglas, H. E. 2009b. Science, Policy, and the Value-Free Ideal. Pittsburgh: University of Pittsburgh Press. Douglas, H. E. 2010. Engagement for Progress: Applied Philosophy of Science in Context. Synthese 177(3): 317–335. Douglas, H. E. 2014. Values in Social Science. In N. Cartwright; E. Montuschi (ed.), Philosophy of Social Science: A New Introduction, p.162–182. Oxford: Oxford University Press. Dupré, J. 2007. Fact and value. In H. Kincaid; J. Dupré; A. Wylie (ed.), Value-Free Science: Ideals or Illusions, p.27–41. Oxford: Oxford University Press. Elliott, K. C.; Richards, T. (ed.). 2017. Exploring Inductive Risk: Case Studies of Values in Science. Oxford: Oxford University Press. Elliott, K. C.; Steel, D. (ed.). 2017. Current Controversies in Values and Science. New York: Routledge. Friedman, M. 1974. Explanation and scientific understanding. Journal of Philosophy 71: 5–19. Glennan, S. 1996. Mechanisms and the nature of causation. Erkenntnis 44(1): 49–71. González, W.J. 2013. Value Ladenness and the Value-Free Ideal in Scientific Research. In C. Luetge (ed.), Handbook of the Philosophical Foundations of Business Ethics, p.1503–1521. Springer: Dordrecht. Green, S.; Jones, N. 2016. Constraint-based reasoning for search and explanation: Strategies for understanding variation and patterns in biology. Dialectica 70(3): 343–374. Gries, S. T.; Newman, J. 2013. Creating and using corpora. In R.J. Podesva; D. Sharma (ed.), Research Methods in Linguistics, p.257–287. Cambridge: Cambridge University Press. Haack, S. 1998. Science as Social? –Yes and No. In Manifesto of a Passionate Moderate, p.104– 122. Chicago: The University of Chicago Press. Haack, S. 2016. Serious Philosophy. Spazio Filosofico 18: 395–407. Hacking, I. 1983. Representing and Intervening: Introductory Topics in the Philosophy of Natural Science. Cambridge: Cambridge University Press. Harding, S. 1986. The Science Question in Feminism. New York: Cornell University Press. Hempel, C. 1965. Aspects of scientific explanation and other essays in the philosophy of science. New York: Free Press. Hooker, C. 2011. Introduction to Philosophy of Complex Systems. In C. Hooker (ed.), Philosophy of Complex Systems, p.841–909. Amsterdam: Elsevier. Huneman, P. 2010. Topological explanations and robustness in biological sciences. Synthese 177: 213–245. Huneman, P. 2018. Outlines of a theory of structural explanation. Philosophical Studies 175(3): 665–702. PRINCIPIA 28(4): 517–540 (2024) 536 Juan Bautista Bengoetxea Intemann, K. 2015. Distinguishing between legitimate and illegitimate values in climate modeling. European Journal for Philosophy of Science 5: 217–232. Keller, E. F. 2005a. The century beyond the gene. Journal of Biosciences 30: 3–10. Keller, E. F. 2005b. Revisiting ‘scale-free’ networks. BioEssays 27: 1060–1068. Kellert, S. H.; Longino, H. E.; Waters, C.K. 2006. Scientific Pluralism. Minneapolis: University of Minnesota Press. Kincaid, H.; Dupré, J.; Wylie, A. (ed.). 2007. Value-Free Science? Ideals and Illusions. Oxford: Oxford University Press. Kitcher, P. 1989. Explanatory unification and the causal structure of the world. In P. Kitcher; W. Salmon (ed.), Scientific explanation, p.410–505. Minneapolis: University of Minnesota Press. Kitcher, P. 2023. What’s the Use of Philosophy?. Oxford: Oxford University Press. Koskinen, I. 2021. Objectivity in contexts: withholding epistemic judgement as a strategy for mitigating collective bias. Synthese 199: 211–225. Koskinen, I.; Rolin, K. 2022. Distinguishing between legitimate and illegitimate roles for values in transdisciplinary research. Studies in History and Philosophy of Science 91: 191–198. Lacey, H. 1997. The Constitutive Value of Science. Principia 1(1): 3–40. Lacey, H. 1999. Is Science Value-Free? Values and Scientific Understanding. London: Routledge. Lacey, H. 2004. Is There a Significant Distinction between Cognitive and Social Values? In P. Machamer; G. Wolters (ed.), Science, Values, and Objectivity, p.24–51. Pittsburgh: University of Pittsburgh Press. Lacey, H. 2011. A imparcialidade da ciência e as responsabilidades dos cientistas. ScientiæStudia 9(3): 487–500. Lacey, H.; Mariconda P.R. 2014. O modelo das interações entre as atividades científicas e os valores. ScientiæStudia 12(4): 643–668. Landes, J. 2020. The variety of evidence thesis and its independence of degrees of Independence. Synthese 198: 10611–10641. Laudan, L. 1984. Science and Values: The Aims of Science and Their Role in Scientific Debate. Berkeley: University of California Press. Longino, H. 1990. Science as Social Knowledge. Cambridge, MA: Princeton University Press. Longino, H. 1996. Cognitive and Non-Cognitive Values in Science: Rethinking the Dichotomy. In L. H. Nelson; J. Nelson (ed.), Feminism, Science, and The Philosophy of Science, p.39–58. Dordrecht: Kluwer. Machamer, P.; Darden, L.; Craver, C. 2000. Thinking about mechanisms. Philosophy of Science 67: 1–25. MacWhinney, B.; Snow, C. 1990. The Child Language Data Exchange System: an update. Journal of Child Language 17(2): 457–472. Mahner, M. 2022. Naturalismo. Trans. Francisco José Mota Poveda [Naturalismus]. Pamplona: Laetoli. Martins Dos Reis, C.R., De Patta Pillar, V. 2018. Valores, estratégias de pesquisa e aplicação do conhecimento: os Campos Sulinos en questão. Principia 22(3): 461–483. McIntyre, L. 2019. The Scientific Attitude: Defending Science from Denial, Fraud, and Pseudoscience. Cambridge, MA: The MIT Press. McMullin, E. 1982. Values in Science. In P.D. Asquith; T. Nickles (ed.), PSA: Proceedings of the 1982 Biennial Meeting of the Philosophy of Science Association (Volume Two, Symposia PRINCIPIA 28(4): 517–540 (2024) Networks, Explanations and Values in Linguistics 537 and Invited Papers), p.3–28. East Lansing, MI: Philosophy of Science Association. McNamara, T. 2006. Validity and values: Inferences and generalizability in language testing. In M. Chalhoub-Deville; C.A. Chapelle; P. Duff (ed.), Inference and Generalizability in Applied Linguistics, p.27–45. Amsterdam: John Benjamins. Moreno, Á.; Suárez, J. 2020. Plurality of explanatory strategies in biology: mechanisms and networks. In W.J. González (ed.), Methodological Prospects for Scientific Research: From Pragmatism to Pluralism, p.141–165. Cham: Springer. Ninio, A. 2006. Language and the leaning curve: A new theory of syntactic development. Oxford: Oxford University Press. Popper, K. R. 1994. The Myth of the Framework. London: Routledge. Potter, E. 2006. Feminism and Philosophy of Science: An Introduction. London: Routledge. Pournari, M. 2008. The Distinction Between Epistemic and Non-Epistemic Values in the Natural Sciences. Science & Education 17: 669–676. Pritchard, D. 2021. Intellectual virtues and the epistemic value of truth. Synthese 198: 5515– 5528. Psillos, S. 2015. Evidence: wanted, alive or dead. Canadian Journal of Philosophy 45: 357– 381. Putnam, H. 2002. The Collapse of the Fact/Value Dichotomy and Other Essays. Cambridge, MA: Harvard University Press. Radder, H. 2003. Technology and Theory in Experimental Science. In The Philosophy of Scientific Experimentation, p.152–173. Pittsburgh: University of Pittsburgh Press. Resnik, D. B.; Elliott, K.C. 2023. Science, Values, and the New Demarcation Problem. Journal for General Philosophy of Science 54: 259–286. https://doi.org/10.1007/s10838-02209633-2 Rezaee, H. S.; Bikaraan-Behesht, H. 2023. Value-Free Ideal is an Epistemic Ideal: An Objection to the Argument from Inductive Risk. Principia 21(1): 137–163. Risjord, M. 2023. Philosophy of Social Science: A Contemporary Introduction. 2nd Edition. New York: Routledge. Rolin, K.H. 2015. Values in Science: The Case of Scientific Collaboration. Philosophy of Science 82(2): 157–177. Rudner, R. 1953. The Scientist Qua Scientist Makes Value Judgments. Philosophy of Science 20(1): 1–6. Salmon, W. 1989. Four Decades of Scientific Explanation. Minneapolis: University of Minnesota Press. Seoane, L.; Solé, R. 2018. The morphospace of language networks. Scientific Reports 8: 10465. Shrader-Frechette, K. 1993. Burying Uncertainty: Risk and the Case Against Geological Disposal of Nuclear Waste. Berkeley: University of California Press. Steel, D. 2010. Epistemic Values and the Argument from Inductive Risk. Philosophy of Science 77: 11–34. Thurner, S.; Hanel, R.; Klimek, P. 2018. Introduction to the Theory of Complex Systems. Oxford: Oxford University Press. Tomasello, M. 2003. Constructing a language: A usage-based theory of language acquisition. Cambridge, MA: Harvard University Press. Van Fraassen, B.C. 1980. The Scientific Image. Oxford: Oxford University Press. PRINCIPIA 28(4): 517–540 (2024) 538 Juan Bautista Bengoetxea Vilhalemm, R. 2016. Chemistry and the problem of pluralism in science: an analysis concerning philosophical and scientific disagreements. Foundations of Chemistry 18: 91–102. Wilholt, T. 2009. Bias and values in scientific research. Studies in History and Philosophy of Science 40(1): 92–101. Zuidema, W.; de Boer. B. 2013. Modelling in the language sciences. In R.J. Podesva; S. Devyani (ed.), Research Methods in Linguistics, p.428–445. Cambridge: Cambridge University Press. Notes 1On the multiple lines of research adoptable in medicine, as a field of high scientific and social interest, see for example Barker and Kitcher 2014, p.153–4. 2Advocates of the ideal of a value-free science might object that the influence of values, if any, is always prior to any research. This is the seminal idea of Reichenbach (1938) when he proposed the distinction between the context of discovery and that of justification, according to which, the mission of the scientist would be to seek and find this justification, to establish a hypothesis well, to justify it, corroborate it and verify it in such a way that non-epistemic values do not intervene in these procedures. This is precisely what is not accepted tout court today. 3Although we follow the four-phase pattern, closer to Douglas and Lacey, other authors, as is the case of Martins Dos Reis & De Patta Pillar (2018, p.464–8), analyze the roles that values exert in five phases of scientific activity: adoption of the research strategy, development of that strategy, cognitive evaluation of hypotheses and theories, spread of scientific results, and application of scientific knowledge. 4Douglas (2009b, p.93) mentions as cognitive values simplicity, scope, consistency, predictive accuracy, fecundity and—the one that interests us most here—explanatory power. Lacey (1999, p.58–60), on the other hand, speaks of empirical adequacy, explanatory and unifying power, encapsulation of possibilities, internal consistency, consonance, support between theories, and the capacity to solve puzzles. 5For example, the fact that some individuals die due to the side effects of a drug wrongly considered safe, or that other individuals die from a disease because they did not have access to a treatment wrongly considered unsafe. 6Putnam’s thick concepts are quite similar to Neurath-Cartwright’s Ballung concepts, although the former are specifically placed in the ethical debate. 7John Dupré (2007) openly claims that thick ethical terms cannot be eliminated from science, at least from certain parts of it. Scientific hypotheses, theories and results concern us just because they affect our human interests, which makes it necessary to express them with a vocabulary flecked with thick ethical concepts. And while it is sometimes possible to translate ‘thickness’ into ‘impartiality’ and ‘neutrality’, the consequence of this always results in some loss of certain human interests, since there are always, more or less surreptitiously, interests of this kind at stake. “If most or all of physics is value free, it is not because physics is science but because most of physics simply doesn’t matter to us”, Dupré (2007, p.31) provocatively says. PRINCIPIA 28(4): 517–540 (2024) Networks, Explanations and Values in Linguistics 539 8Anyway, this perspective is not unanimous in the philosophy of science. Betz (2013), for example, demands more precision in the treatment of the role of (non-epistemic) values in the context of a criticism of a science without values. According to him, this criticism is tantamount to just flogging a dead horse (see Pritchard 2021, Pournari 2008). 9As McIntyre (2019, p.230) writes, “while it is true that we should not let our hopes, wishes, beliefs, and ‘values’ colour our inquiry into the ‘facts’ about human behaviour, this does not mean that values are unimportant. Indeed, as it turns out, our commitment to the scientific attitude is an essential value in conducting scientific inquiry”. 10Such a strategy would be the explanatory one, for example, aimed at obtaining its own results thanks to “a dense interwoven net of scientific constructions that constrain and support each other (...) ideas, concepts, theories, experiments, measures, middle-level principles, models, methods of inference, research traditions, data, narratives” (Cartwright 2022, p.12, 20; see also Cartwright et al. 2022), and so on. Among them, we conclude that also values (cognitive, epistemic and non-epistemic) must be introduced. 11We follow the standard format of the four phases and leave out one phase that Lacey and Mariconda (2014, p.645) do take into account: spreading scientific results. We think that this and ‘our’ Ph-4 can be unified. In any case, this does not affect the goal of our article. 12Mahner (2022, p.68) qualifies the slight difference between the concept of data and that of evidence. According to him, data are the result (expressed linguistically) of an empirical operation (observation, measurement, experiment), but they are not evidences. As such, data do not say anything by themselves. A datum dbecomes evidence only in relation to a hypothesis hor theory t. This is important in many epistemological contexts, but here we will treat both concepts equivalently because the slight difference does not interfere with our working thesis. 13For cases of weaker relationships such as isomorphism or homomorphism, see Bueno (2011, p.253–4). 14Deulofeu, Suárez and Pérez-Cervera (2021, p.2010–1) present a structural, not mechanistic, explanation of the stability behavior of the microbiome. They point out that, although Coyte¸ Schluter and Foster (2015) appear to provide a model of the mechanism defined in terms of entities, activities, and their organization, the details of the causal pathway that cause the stable behavior are actually diluted in the complexity of a mathematical analysis. 15In mathematics and physics, a small-world network is a kind of network for which most of the nodes are not neighbors of each other, although, nevertheless, almost all of them can be reached from any source node through a relatively short number of steps between them (Barabási 2002, p.53). That is, the construction of a small-world network depends on (1) the existence of a small-world phenomenon as such (any two nodes in the network communicate over a relatively small intermediate node path (small number of nodes), since the maximum distance between two nodes grows logarithmically along with the total number of nodes in the network), and (2) that it possesses high clustering coefficients or, in other words, the indication that, if two nodes are not directly connected to each other, there is a high probability that they connect through the intervention of other nodes. Therefore, it is a property of a small world that it presents networks in which, despite the existence of a large number of nodes, it is possible to find short communication paths between them. On the other hand, a scale-free network is a specific kind of complex network. In a scale-free network, some nodes are highly connected, i.e., they have a large number of links to other nodes, PRINCIPIA 28(4): 517–540 (2024) 540 Juan Bautista Bengoetxea although the degree of connection of almost all other nodes is quite low. Barabási (2002, p.58) points out that the Web, for example, does not have a distribution of the usual degree of connectivity. Instead, a few nodes, which are called hubs, are much more connected than the rest. An example is the case of networks of neurons in organisms with nervous systems. In these, it is quite normal to use very frequently a fraction of the neurons while leaving most of the rest of the neurons barely utilized. 16CHILDES (Child Language Data Exchange System) is a corpus (as of 2015 it had contents– transcripts, audio, and video—in 26 languages from 230 different corpora) to serve as a central repository for data of first language acquisition. See MacWhinney and Snow (1990). 17A SAN software is a high-speed local network that contains multiple sets of block storage devices. Among some of its features, SAN can consolidate storage devices (eases management), efficiently pools storage devices, improves performance by reducing storage latency, improves application availability through redundancy, simplifies data backup (no server interaction), and supports multiple kinds of storage devices. 18Douglas (2009b, p.108–12) analyzes the indirect role of values (cognitive and social) in the case of diethylstilbestrol (DES) (a compound that operated as an estrogen). According to her, the explanatory value in this case only plays an orienting role in decision making. However, as Coyte, Schluter and Foster (2015), Deulofeu, Suárez and Pérez-Cervera (2021), and Moreno and Suárez (2022) point out, there are other cases in which the explanation adopts more complex forms, as in the case of the microdiome. We think that these more complex explanations, which are sometimes of a structural-mathematical nature or, also, integrated with a mechanistic explanation format (Deulofeu; Suárez; Pérez-Cervera 2021, p.2022), do play a direct role, as we have tried to show in the case of experimental linguistics. Acknowledgments Grant PID2020-113449GB-I00, funded by MICIU/AEI/10.13039/501100011033/, and grant PID2023-147251NB-I00 funded by MICIU/AEI/10.13039/501100011033 and by ERDF/EU. I would also like to thank the two anonymous referees for their contribution to the improvement of this paper. PRINCIPIA 28(4): 517–540 (2024)