Minds & Machines https://doi.org/10.1007/s11023-018-9463-8
Syntactical Informational Structural Realism Majid Davoody Beni1
Received: 12 October 2017 / Accepted: 3 April 2018 Springer Science+Business Media B.V., part of Springer Nature 2018
Abstract Luciano Floridi’s informational structural realism (ISR) takes a constructionist attitude towards the problems of epistemology and metaphysics, but the question of the nature of the semantical component of his view remains vexing. In this paper, I propose to dispense with the semantical component of ISR completely. I outline a Syntactical version of ISR (SISR for short). The unified entropy-based framework of information has been adopted as the groundwork of SISR. To establish its realist component, SISR should be able to dissolve the latching problem. We have to be able to account for the informational structures–reality relationship in the absence of the standard semantical resources. The paper offers a pragmatic solution to the latching problem. I also take pains to account for the naturalistic plausibility of this solution by grounding it in the recent computational neuroscience of the predictive coding and the free energy principle. Keywords Informational structural realism Free energy principle The unified entropy-based framework of information Predictive processing Syntactical informational structural realism
1 Introduction The significance of semantical breakthroughs in the mid-twentieth-century could hardly be exaggerated. Among other things, the semantical turn had enriched the philosophy of science of Tarski’s contemporaries. Later, the articulation of model theory resulted in further semantical reforms in the philosophy of science. & Majid Davoody Beni
[email protected] 1
Philosophy of Science Group, Department of Management, Science and Technology, Amirkabir University of Technology, No. 424, Hafez Street, PO Box 3313-15875, Tehran, Iran
123
M. D. Beni
Accordingly, semantical notions of ‘designation’ and ‘reference’, as well as structural relations of ‘isomorphism’ and ‘partial isomorphism’, were invoked to fructify the semantic component of some advanced forms of scientific (structural) realism. The present paper is concerned with informational structural realism (ISR) and its semantical aspects. Luciano Floridi’s (2008, 2009) version of ISR elicits a constructionist attitude towards the questions of metaphysics and epistemology. However, it is not always easy to connect various aspects of Floridi’s work. The connection between ISR and Floridi’s several attempts at offering a theory of semantic information is such a case. The semantic component of ISR may draw on Floridi’s (2004) Strongly Semantical Theory of Information (SSTI). However, there may be reasons to suspect that SSTI is committed to representationalism. Adriaans’ (2010) critical analysis of Floridi’s theory of semantic information is based on such suspicions. This can indicate that Floridi’s version of ISR is inconsistent—because the representationalist or mimetic tendency does not mesh nicely with the general constructionist tendency of ISR. Under the circumstances, it may be possible to underscore the constructionist undertone of Floridi’s philosophy of information (and ISR). On several occasions, Floridi (2011a, 2014, 2016) spelt out the constructionist commitments of his theory of semantic information. On such grounds, it may be possible to argue that Floridi’s view on semantic information is not committed to the orthodox mimetic theories. But then again, we have to also substantiate the plausibility of the constructionist stance in the face of possible criticisms. The more radical option is to dispense with the semantic component of ISR completely and try to defend a Syntactical version of ISR (SISR for short). While these two paths to defending the intactness of ISR will meet again eventually, I will mainly focus on treading the latter path in this paper. The paper consists of seven sections. In the second section, I canvass the relation between SR and the problem of semantics. The third section concerns the connection between ISR and SSTI. In this section, I also allude to Adriaans’ (2010) reservation about the (mimetic) connotations of SSTI. In the fourth section, I outline SISR. In the fifth section, I offer a pragmatic solution to the latching problem. Despite being committed to syntax, SISR, too, is a structuralist version of realism. So, to defend SISR convincingly, we have to be able to explain how SISR could account for the relation between the informational structures on the one hand, and the enduring objective world on the other hand. I call this problem ‘the latching problem’, and I argue that SISR can deal with the latching problem on the basis of pragmatics instead of semantics. The fifth section outlines this pragmatic solution. The sixth section contains a more expansive reply. This reply draws on the resources of recent computational neuroscience and theories of the brain’s predictive coding. This move is supposed to provide a scientifically informed basis for reinforcing the pragmatic solution that I offer in this paper. I will conclude the paper by remarking that SISR’s scientifically informed pragmatic reply to the latching problem is consistent with a possible extension of Floridi’s remarks on the constructionist undertone of the semantic component of ISR.
123
Syntactical Informational Structural Realism
2 Structural Realism and Informational Structural Realism Structural Realism (SR) is an (epistemic/ontic) realist improvement upon the Semantic View of Theories (SVT). SVT emerged as a reaction to the so-called syntactical view [or Received View of Theories (RVT)] which had been advocated by some logical empiricists. Despite its reputation, RVT had absorbed the essence of Tarski’s semantical breakthroughs in logic almost as soon as they had been available (Tarski 1944; Carnap 1942). Even so, by the 1960s RVT began to lose grounds to the set/model-theoretic approach of Suppes and van Fraassen. Modeltheoretical approach (or SVT) purported to dispense with the meta-linguistic formulations of the scientific theories and embarked on direct designation and specification of the intended family of models of the theory (in the modeltheoretical sense) (Suppes 1962, 1967; van Fraassen 1980). According to SVT, the structures of scientific theories are identifiable with the families of models (e.g. as state spaces or set-theoretical structures) that stand in a mapping relation to them. SVT was supposed to provide a better understanding of the nature of scientific theories and scientific activity (Suppe 1998). The structural realists re-inflated the epistemological and ontological commitments of SVT. According to different formulations, SR holds that scientific theories are informative about the structure of the world (Worrall 1989), or/and it asserts that the structure is all there is (French and Ladyman 2003, 2011). Different pieces of evidence from the history of science and ontology of modern physics have been conjured in support of SR. SR has been developed into epistemic, ontic, eliminativist, and non-eliminativist varieties. I have to emphasise the point that SR does rely heavily on the representationalist, formal semantics of model theory and the formal relations of isomorphism and partial isomorphism. It is true that some structural realists have attempted enriching the semantical notion of truth pragmatically (da Costa et al. 1998; da Costa and French 2003). But even da Costa and French’s notion of ‘‘pragmatic truth’’ has been offered as an extension of Tarskian semantics. All in all, it has been assumed that the realist component of SR is connected with representationalism. The structural realists unanimously claim that scientific theories represent the mind-independent structure of reality almost accurately (Worrall 1989; French 2014, 2015). Informational SR (or ISR) offers to reconcile the ontic version of SR to its epistemic version. ISR has been outlined as an extension of Luciano Floridi’s attempt at providing a foil to both digital and analogue informational ontologies (Floridi 2009). Floridi defined ISR in the following manner: Explanatorily, instrumentally and predictively successful models (especially, but not only, those propounded by scientific theories) at a given LoA [i.e. Level of Abstraction] can be, in the best circumstances, increasingly informative about the relations that obtain between the (possibly subobservable) informational objects that constitute the system under investigation (through the observable phenomena). (Floridi 2008, 161:240–41) The point that is worth underlining is that ISR diverges from the mimetic goals of orthodox ontic SR, and it leans towards constructionism. The constructionist
123
M. D. Beni
tendency of ISR has roots in Kantian origins of Floridi’s theory as well as his insights concerning the role of technology in manufacturing knowledge (Floridi 2011b, 2017). The semantic component of Floridi’s ISR seems to be in harmony with the general constructionist attitude of ISR. For example, in the conclusion of his paper, Floridi briefly hinted that his views on reference and designation are different from what one expects from the traditional representational semantics. According to Floridi, instead of describing a precise picture or representation of the environment, we interact with the environment as ‘‘a resource for our semantic tasks’’, and interrogate it through ‘‘experience, tests and experiments’’ (Floridi 2008, 161:249). This brief hint reveals that Floridi’s conception of semantics is different to the typical representationalist accounts of the theories-world relationship. Floridi’s view on semantics is in line with his constructionist conception of reality, according to which: Reality in itself is not a source but a resource for knowledge. Structural objects (clusters of data as relational entities) work epistemologically like constraining affordances: they allow or invite certain constructs (they are affordances for the information system […] that elaborates them) and resist or impede some others (they are constraints for the same system), depending on the interaction with, and the nature of, the information system that processes them’’ (Floridi 2009, 176). I have to add that, despite Floridi’s expressed commitment to constructionism, the connection between ISR and Floridi’s work on semantic information has not been spelt out clearly. Floridi’s two papers on structural realism (Floridi 2008, 2009) are almost silent about the nature or technical properties of the semantic component of ISR. Can degrees of accuracy and informativeness be directly used to assess how a model of reality is related to the reality it is a model of? A constructionist may deny that we need to assess the relation between models and reality directly. But then again, we have to be able to defend plausibility of the constructionist basis of this reply.
3 Strongly Semantical Notion of Information, and a Critical Analysis It might be possible to assume that SSTI underlies the semantic component of ISR. This assumption is based on some general hints, e.g., (Floridi 2009, 167, and footnote 24). Below, I shall argue that even if we accept this assumption, new problems will raise their head. Floridi’s (2004) SSTI was based on truth-values instead of probability distributions which had been at issue in the theories of weakly semantical information (e.g. Elias, Carnap, and Bar-Hillel 1954). Weakly Semantical Theories of Information (WSTI) had been liable to the notorious Bar-Hillel-Carnap Paradox. The paradox draws attention to a seemingly self-contradictory aspect of the probabilistic (weakly semantical) theories of information. It holds that:
123
Syntactical Informational Structural Realism
[i]t might perhaps, at first, seem strange that a self-contradictory sentence, hence one which no ideal receiver would accept, is regarded as carrying with it the most inclusive information. It should, however, be emphasized that semantic information is here not meant as implying truth. A false sentence which happens to say much is thereby highly informative in our sense. Whether the information it carries is true or false, scientifically valuable or not, and so forth, does not concern us. A self-contradictory sentence asserts too much; it is too informative to be true. (quoted from Floridi 2004, 197). Floridi’s SSTI dispenses with the probabilistic account of semantic information and thus avoids Bar-Hillel-Carnap Paradox. SSTI assigns truth-values on the basis of discrepancy with respect to a given situation, and it lines up with the ordinary general sense of semantic information. Despite its evident virtues, it is not obvious that SSTI contributes to the constructionist goals of Floridi’s ISR (given the connection between SSTI and ISR). Below, I shall unpack the reasons for this scepticism. According to SSTI, the degree of discrepancy (as want of agreement) between ri and a given state of the world w can be calculated as the degree # of semantic deviation of ri from the uniquely determinate state w in which E is (||- W#r). r stands for ‘‘infon’’, and it refers to discrete items of factual information. Infons are qualifiable in principle as either true or false, irrespective of their semiotic code and physical implementation. The amount of informativeness of each ri is evaluated as a function of the alethic value possessed by ri (Floridi 2005, 205). The proposal has been extended to include degrees of inaccuracies and informativeness, the quantity of misinformation and the degree of disinformation. On such grounds, Floridi defined factual or semantic information as well-formed, meaningful and truthful data. Note that Floridi explicated that information about a situation encapsulates truth, and the degree of informativeness is defined in relation to a factual situation in the world. This definition has the appearances of being consistent with representationalist semantics, which defines reference and designation in terms of correspondence between symbols and states of affairs. It could be observed that the way of a mimetic interpretation is not in (Floridi 2004). Such an interpretation inspired Pieter Adriaan’s critical analysis of SSTI. Although Adriaans agreed that Shannon’s theory does not account for the meaning of information, he pointed out that this is not enough reason ‘‘to develop a special [form] of semantic information in order to repair this deficit’’ (Adriaans 2010, 51). This is because there are a number of extensions of information theory that are capable of dealing with the elements of meaning (namely different kinds of WSTI, or even mathematical theories). Adriaan argued that Floridi had begged the question of the meaningfulness of information, in order to show Shannon’s theory is not centred on the appropriate notion of information (Adriaans 2010, 43). In contrast, Adriaans suggested that even without engaging the issue of meaning and semantics, Mathematical Theory of Communication (MTC for short) could ground the ‘‘central notions of theory of knowledge with unprecedented mathematical rigor’’ (47).
123
M. D. Beni
I have to add that Adriaans’ greatest dissatisfaction with Floridi’s proposal seems to be concerning SSTI’s philosophical consequences. He complained that Floridi had tried to convert information theory to a servant of orthodox epistemology. But according to Adriaans, the philosophy of information is a competitor of the traditional epistemology, rather than its subsidiary. This means that the philosophy of information could deal with the problem of knowledge, even without conforming to the framework of orthodox epistemology. So, although Adriaans reasons for being disagreeable to SSTI are multiple, it seems that his main reservation was motivated by the point that Floridi’s SSTI could be easily understood as an orthodox mimetic view. It is possible to demonstrate that Adriaans’ analysis overlooks the constructionist tendency of Floridi’s philosophy of information. The constructionist tendency has been expressed on a number of different occasions. If we take the constructionist tendency of Floridi’s philosophy of information into account, we can see that Floridi did not intend to degrade the philosophy of information to a servant of traditional epistemology (at all). For example, we can consider Floridi’s (2011a) Correctness Theory of Truth (CTT) which diverges from Tarski’s theory in taking a pragmatic (as opposed to model-theoretic) and hence an exogenous turn. According to CTT, knowledge encapsulates truth, which encapsulates semantic information. Something is semantic information if it is true, and if its correctness could be verified and validated1 (in the computer science’ sense) (159–160). The notion of truth could be defined in terms of the reply to a query within a given model and a given context, at a level of abstraction, and for a specific purpose. All of these bestow a constructionist character upon CTT. Moreover, CTT holds that something counts as semantic information only with respect to certain kind of informee, as an embodied and embedded, creative agent, who actively interacts with reality so as to shape and build it. The orthodox mimetic forms of epistemology and semantics are not usually agent-based. Perhaps it is also worth mentioning that CTT is inspired by ‘‘design-pattern’’ technique in software engineering. The technique is based on specifying the abstract features of a design structure at the meta-level. The design structures are meta-level reusable solutions to commonly occurring problems in the construction of an artefact (ibid, 152). In all, CTT holds that the necessary and sufficient correctness-maker is the whole complex construct which is represented by the dynamical configuration of the entire distributed system. This indicates that Floridi’s conception of semantics diverges from representationalism. It also indicates that Floridi is not concerned with the orthodox epistemology.2 To be clear, it is true that CTT is originally a theory of truth, but in developing the theory does sketch the bigger picture and the connection with knowledge. 1
Verification and validation could be defined on the basis of checking whether we are constructing (or have constructed) what we have (or had) planned to construct checking whether we are constructing what is required.
2
The constructionist approach to truth and knowledge resurfaced more recently in (Floridi 2016), where he asserted that on some occasions, a proposition, a message, or some information qualifies as being knowledge (and truthful) not just in itself but relationally, with respect to an informational agent. He also demonstrated that there are cases in which the poietic (constructive) intervention on a system determines the truth of the model of that system.
123
Syntactical Informational Structural Realism
Given Floridi’s emphasis on the constructionist essence of ISR, construing the semantic component of ISR in terms of representationalism would be undesirable. But a radical departure from the orthodox semantics could be confusing, too. Pragmatics is concerned with the role of context in the formation of (the speaker) meaning, whereas semantics is about reference and designation. Floridi’s constructionist semantics borders on pragmatics. That is to say, Floridi’s constructionist view on semantic information takes the issues of the context, agent, and purpose into consideration. As Floridi acknowledged, these are exogenous and pragmatic factors. What amount of pragmatics can be absorbed by semantics, before the field of semantics loses its semantical character completely? To what extent semantics can be imbued with pragmatic factors, before the pragmatic factors begin to suppress the semantic character of the theory? It may be possible to find the right combination of the pragmatic and semantic factors and outline an optimised version of constructionist semantics. But I do not directly engage in establishing the plausibility of constructionism in this paper (although I do not deny its plausibility either). In the remainder of the paper, I shall present a refined version of ISR (or an interpretation of it) that is not burdened by the complications caused by the assumption of the existence of a semantic component at all. In a nutshell, I shall replace ISR with a Syntactical version of ISR (SISR, for short). There is, however, a serious obstacle in the way of developing SISR. Unless I could outline a solution to the problem of theories-reality relationship, the syntactical version of informational structuralism would succumb to syntactical instrumentalism. But according to my theory, SISR is still a version of structural realism, and I shall retain the ontological component despite dispensing with the semantic one. To fulfil this goal, I shall deal with the ‘latching problem’ (i.e., the problem of the relationship between the informational structures and reality) in pragmatic terms instead of semantic ones. In the remainder of this paper, I shall firstly spell out SISR and then resolve the latching problem without invoking semantics.
4 A Syntactical Proposal SISR sheds the semantic component of ISR. While I think such syntactical improvement is essentially consistent with Floridi’s version of ISR, I do not insist on attributing my point of view to Floridi. Therefore, I present SISR as an independent variety of informational structuralism that could stand on its own feet. Let us begin with a clarification about the syntactical nature of my proposal. Truth is traditionally a semantic notion—specified at the level of the metalanguage. But the solution that I will develop here won’t hinge on the notion of truth. SISR is syntactical in a general sense, because it does not rely on the semantical notion of truth (in either model-theoretic or information-theoretic sense). However, it is not syntactical in the pre-semantical sense of, say (Carnap 1937), because it is not committed to the syntax in the sense that is at issue in predicate logic or model theory. SISR is syntactical in another specific sense, i.e., in the sense of Shannon’s Mathematical Theory of Communication (or MTC). SISR is
123
M. D. Beni
syntactical in the sense of not taking into account what strings of symbols drawn from a given alphabet mean.3 So, the proposal that I develop here is syntactical in two senses. It does not invoke meta-linguistic truth (this is the general sense). Neither does it rely on the notion of meaningful and truthful semantic data which is at issue in semantic information. Instead, we could understand theories in terms of the unified entropy-based informational framework. Let me elaborate. MTC is concerned with the questions of the ultimate level of data compression and the ultimate rate of data transmission. MTC defines the measure of the average uncertainty in the message when the signal is known (Shannon and Weaver 1949). When stated technically, Shannon’s theory holds that the amount of ‘disorder’ H(X) in a collection of messages is expressible in terms of a probability distribution P over the set of messages. That is to say, the communication entropy of X for a set of messages xi = (I = 1, … n) is: X Pðxi ÞlogPðxi Þ H ðXÞ ¼ i¼1;n
Note that MTC does not give rise to Bar–Hillel–Carnap Paradox, because MTC does not aim to make truth-values supervene on semantics of information. Adriaans argued that the notion of Shannon information could be combined with the notions of Gibbs entropy and Kolmogorov complexity. The combination could amount to a unified framework of an entropy-based account of information (Adriaans 2010, Section 2). Historically, Shannon had derived his notion of communication entropy out of Gibbs entropy. Gibbs’ notion of entropy consists in a measure of the amount of ‘disorder’ S in a closed system of microstates in equilibrium in terms of the probability distribution of the energies of the system. Assuming that pi is the probability of the occurrence of a microstate during fluctuation, the entropy of the system can be defined as: X S¼ pi ln pi i
Gibbs’ theory defines the distance between the entropy of the actual system and maximal entropy in terms of ‘‘free energy’’4 (Adriaans 2010, 44). The notion of free energy is rather important, and we will return to it in the sixth section of this paper. But right now, I have to emphasise the point that the unified entropy-based framework of information does not accommodate semantic information in the sense that Floridi’s SSTI does. Even so, at least according to Adriaans, the unified entropy-based framework could be adopted as a ‘‘general framework to study human cognition and methodology of science’’ (44). I welcome the proposal, and in the sixth section of this paper, I will elaborate on the connection between this unified framework and schemas of human cognition. With an eye to this 3
This is different from the first sense of ‘syntactical’. For example, the theory of semantic information of Bar-Hillel and Carnap is semantic in the second sense, but syntactic in the first sense (it is based on statedescriptions). This point has brought to my attention by one of the referees of this journal.
4
Gibbs entropy is derived out of (F ¼ T ln Z) where F is free energy, T is equal to fixed temperature, P and Z is partition function equal to i ei =T :
123
Syntactical Informational Structural Realism
advantageous point, I adopt the unified entropy-based informational framework as the groundwork for developing SISR. SISR holds that: Explanatorily, instrumentally and predictively successful models (especially, but not only, those propounded by scientific theories) are in the best circumstances, increasingly informative about the relations that obtain between the (possibly sub-observable) informational objects that could be identified within the unified entropy-based general informational framework. SISR is quite close to the stance that Floridi has defended in his articulation of ISR. The only difference is that SISR does not commit itself to a specific theory of semantics. SISR replaces semantics with pragmatics. Despite this difference, SISR is still a form of informational structuralism, because it indicates that scientific models are informative about relations that obtain between informational objects that could be presented within the unified entropy-based general informational framework. So, SISR is a structuralist theory. To be more precise, both elements of SR, i.e., commitments to structuralism and realism, are retained by SISR. Below, I shall elaborate on what it means for a form of structural realism in the philosophy of science to be syntactical. Syntactical Realism is a par excellence form of SR. This is because a syntactical approach does not need to make any commitment to the referents of the theoretical terms that feature in the structure of a scientific theory at all. So, it relinquishes the content and retains the form. Syntax is nothing but the structure of the theory. This construal lines up with the usual sense of ‘syntax’. This is because the syntax is mainly concerned with studying the structure of (an artificial or natural) language. It is not concerned with designation of the terms of language. Therefore, Syntactical Realism could easily identify with Structural Realism. To substantiate this point, I can appeal to the authority of no less significant a figure than the founder of the contemporary form of SR. For, John Worrall, too, identifies Syntactical Realism with SR. He asserted that a form of realism that could be retained in the face of the antirealist argument (i.e., argument form discontinuity of the history of science) consists of a position that ‘‘might be called structural or syntactic realism’’ (Worrall 1989, 112 original emphasis). In the same vein, Worrall drew on the preceding work of Poincare, to assert that Poincare ‘‘used the example of the switch from Fresnel to Maxwell to argue for a general sort of syntactic or structural realism quite different from the anti-realist instrumentalism which is often attributed to him’’ (Worrall 1989, 117 original emphasis). Worrall’s theory endeavours to account for the continuity of the history of science in terms of syntax or structure, not of content. And for Worrall, too, Syntactical Realism is Structural Realism ipso facto. SISR is a version of Syntactical Realism which offers to dispense with semantics. To be clear, while Worrall’s original presentation of SR uses Syntactical Realism as a synonym for SR, orthodox versions of SR do not dispense with semantics in the radical way that SISR does. I just mentioned Worrall’s work to show that syntactical realism is a par excellence form of SR. However, as we have already seen, other forms of SR include semantical commitments. So, it is not the case that every version of SR is also a version of Syntactical Realism, in the narrow sense that is at issue in this paper.
123
M. D. Beni
In my view, SR may benefit from dispensing with the semantical components. For one thing, dispensing with semantics make it possible to go beyond mimeticconstructionist debate. As we have already seen, Floridi and the majority of structural realists cannot see eye to eye with regard to the mimetic/constructionist tendency of semantics of SR. I suspect that from a Kantian perspective, the dichotomy can be understood as a new form of antinomy (an antinomy is a contradiction that follows from the attempt at grasping the nature of the transcendental reality). I understand that Floridi presented constructionism as a Kantian reaction to representationalism. However, notice that even representationalism (or at least some versions of it) could be construed along the lines of a Kantian theory too (for a detailed discussion see Beni 2016). In all, I think that the debate between representationalism versus constructionism could not be satisfactorily concluded in the philosophy of science. Be that as it may, as SISR does not include a semantic component, it does not need to engage in the debate about representationalism versus constructionism. It should be added that SISR’s emphasis on the role of syntax does not mean to indicate that semantics does not matter. Rather, the point is that we do not need to commit ourselves to a particular theory of semantics. Instead, we may try to account for the theories-world relationship without invoking any formal theory of semantics at all. And as I shall argue in the next sections, the borders between informal semantics and pragmatics are fluid. SISR is not an instrumentalist theory. It does not take syntactical-informational structures as mere useful regimentations. SISR is committed to a Kantian version of realism. To retain the realist core of SISR, I should be able to account for the relationship between the informational structures and reality. That is to say, I should be able to deal with the latching problem, and I should do so without drawing on the semantical resources. Otherwise, SISR would fail to be a realist theory, and it would succumb to syntactical instrumentalism. The question is, how does SISR face the latching problem? The short answer to the question is ‘pragmatically’. Of course, it is generally assumed that the latching problem has a semantical solution. But I argue that it is possible to find a pragmatic reply that bypasses the borders of (both modeltheoretic and information-theoretic) existing forms of semantics. The longer answer is that there are scientific accounts of free-energy principle which could be extremely helpful in developing a solution to the latching problem, in way that is agreeable to both Floridi and Adriaans. The longer answer includes the naturalist core of the proposed strategy. The solution to the latching problem is naturalistic in the sense of being informed by the theoretical and empirical psychology. In this sense, my solution makes an improvement upon Floridi’s attempt at articulating semantics on the basis of software engineering and computer sciences.5 In the next two sections, I will unfold both the short reply and the long one. 5
Although I do not endorse a fully human constructionist stance, I do not think that the proposal that I develop here is at odds with Floridi’s (2017) remarks on the significance of the constructionist mechanisms of the formation of knowledge. This is because I do not assert that construction would be reducible to a natural process without information loss. The main difference is that while Floridi draws on computer science to articulate his conception of semantics, I inform my solution by drawing on the resources of empirical psychology and computational neuroscience. Also, notice that that Floridi simply
123
Syntactical Informational Structural Realism
5 Latching onto the World: The Short Reply In the wake of the development of Tarskian semantics, it has been assumed that it is the job of semantics to deal with the latching problem. The advocates of proto-SVT, SVT-theorists, and structural realists shared this insight. I offer to dispense with semantics and resolve the latching problem on the basis of pragmatics. This proposal is compatible with the outcome of Adriaans’ critical assessment. In the conclusion of his (2010) paper, Adriaans remarked that ‘‘syntactically well formed, meaningful, and truthful’’ are conditions of information being real information. But ‘‘although these things are […] pragmatically true, they do not force us to develop a specific theory of semantic information’’ (54). This pragmatic proposal is conformable with SR. There are even precedent cases, which could be mentioned here to substantiate this claim. Wolfgang Stegmuller, for instance, offered a pragmatic solution to the latching problem too (Stegmu¨ller 1979). According to Stegmu¨ller (1979), pragmatics is essentially an informal theory of reference or designation. Stegmuller claimed that SVT-theorists’ reliance on the isomorphism between models allows them to deal with inter-theoretical relations. But to deal with the relations between physical theory and what is outside (in the external world), the structural realists have to go beyond formal semantics. The instrumentalist versions of SVT (as being developed by van Fraassen and Suppes) did not aim to deal with the infra-theoretical relations. Structural realists (e.g. French and Ladyman 1999), on the other hand, aimed to account for the theoriesreality relationship. To do so, the structural realists are compelled to go beyond isomorphic relations, and they have to enrich the representational relations with pragmatic factors (da Costa et al. 1998; da Costa and French 2003). While assessment of this pragmatic intervention is beyond the scope of the present paper, I just point out that structural realists’ attempt at enriching Tarskian semantics confirm Stegmuller’s opinion about the poverty of Tarskian semantics or modeltheoretic formal semantics in dealing with the infra-theoretical relations. Advocates of the orthodox SR see formal semantics as the right venue for finding a solution to the latching problem, but Stegmuller openly defended a pragmatic solution to the problem. It is Stegmuller’s definition of pragmatics as formal semantics is not precise. It is also true that the borders between semantics and pragmatics are fluid. But this does not mean that semantics (in the model-theoretic sense) and pragmatics are the same. Semantics is concerned with the formal theories of designation, denotation, reference, and truth, whereas pragmatics includes what is connected to persons, knowledge situations, and the mechanisms of confirmation of a theory and test procedures (Stegmu¨ller 1979, chapter 5). Let us recap. It is possible to solve the latching problem by taking the path of pragmatics. The proposal is not unprecedented in the structural realist literature. Of
Footnote 5 continued offered a plea for non-naturalism. Therefore, I do not need to engage a debate over the correctness of the naturalist stance, even if my conception of naturalism were at odds with Floridi’s constructionism. But the mentioned stances are not at odds.
123
M. D. Beni
course, the existence of the precedent cases does not guarantee the plausibility of a proposal. We have to develop this short reply into a more extensive one.
6 Latching onto the World: The Long Reply In this section, I shall support my pragmatic solution to the latching problem on the basis of some theories of recent computational neuroscience. I draw on the theories of brain’s predictive coding and free energy principle. The brains of human beings generate hypotheses about the world. The brains are also capable of decreasing the discrepancy between their models and the world through predictive error minimization mechanisms. Scientific theories are the sophisticated extensions of the brain’s rudimentary hypotheses about the world, and scientists use their brain to coalesce information about the world into the corpus of scientific knowledge. Therefore, the scientific account of the brain-world connection could underlie the account of the theories-world relationship6 in the philosophy of science. Alluding to the foundational part of the cognitive processes in the constitution of the scientific theories does not need to undermine the role of social and historical factors. The relation between social and historical aspects of scientific thinking and the underlying cognitive processes could be studied in its own right. But in this paper, I disregard the role of the social and historical factors for the sake of simplicity. Instead, I rely on a neurocomputational account of the brain-world relationship to find a scientific foothold for supporting the pragmatic solution to the latching problem. Before delving into details about the psychological mechanisms that underpin my reply to the latching problem, I have to make a quick clarification. The proposal that I will develop in the following pages is focused on the nature of human cognition, and it does not say much about the nature of reality per se. This may indicate that my proposal is not related to the original problem of (informational) structural realism. I briefly clarify my stance. The version of SR that I defend here comes with a Kantian flavour. It presumes that our access to reality is indirect and mediated by the cognitive system’s action-oriented mechanisms. That is to say, the access to the world is mediated by the epistemic agent’s cognitive schematism, and there is no unmediated access to the nature of reality per se. Even so, focusing on the nature of the cognition of the human beings—as embodied epistemic agents who produce scientific theories and verify them through interacting with the world—leads to a viable (pragmatic) account of the nature of the scientific theories-reality relationship. This is more or less compatible with the Kantian undertone of Floridi’s philosophy of information. According to Floridi’s (2014) restatement of the Kantian theme, the perceptual information about the world is the world. Our perceptions are informative about the nature (or at least structure) of the external causes that affect our sense-organs, and how such an effect manifests itself depends quite essentially on nature of the 6
This could be worded as the ‘mind-world relationship’ too. But as this paper is concerned with the philosophy of science, I articulate the problem in a way that underscores what is important from the perspective of the philosophy of science.
123
Syntactical Informational Structural Realism
perceptual systems on which the causes operate (see Floridi 2014). Floridi, too, drew on a Helmholtzian extension of Kantian realism, so as to establish the ontic component of his view. In this respect, I am in complete agreement with Floridi. For, my solution to the latching problem is based on a theory of the brain’s predictive coding. And according to some interpretations (Friston and Stephan 2007; Hohwy 2013), the predictive coding theory that underpins my account of theories-world relationship is an updated version of Helmholtzian statistical theory of perception (see the next subsection). My neurologically informed solution to the latching problem is loyal to this form of realism. That is to say, my realism is tailored to the neurological account of the nature of the perceptual system on which the causes operate. We can infer the structure of the external causes, i.e., reality, by focusing on the study of the nature of perceptual systems and their (action-oriented) representational capacities. Therefore, it is true that SISR is silent about the nature of reality per se. Instead of focusing on the real nature of the world in itself, SISR considers reality as the source of information that could be exploited (though not exhausted) by the perceptual mechanisms of the organic cognitive systems which enact in the world (as the source of information). The emerging conception of realism is modest and scientifically informed. It builds upon the scientific accounts of how the cognitive agents capture the essential structure of reality. It also assumes that patterns of the brain’s information processing reveal the essential structure of reality as the source of information. Through dynamical interplay with the world’s windows of affordance, the brain succeeds at decreasing the discrepancy between its models and capturing the patterns of regularities of the external world. Therefore, SISR accommodates both epistemological and ontological components, despite assuming that access to the source of information (or reality) is mediated by the brain’s action-oriented perceptual mechanisms. I proceed to provide some further details. 6.1 Predictive Processing The brain’s Predictive Processing theory (PP) is a flourishing theory of computational neuroscience, and it seems to be gaining grounds steadily. The formal and experimental bases of PP are fairly solid, and the theory promises to unify the fields of perception, memory, cognition and action (Rao and Ballard 1999; Srinivasan et al. 1982; Friston and Price 2001; Frith 2007; Friston 2012). More recently, philosophers of cognitive science spotted the capacity of PP for informing the philosophical discussions about mental representation and the brain-world relationship (Clark 2012, 2016a; Hohwy 2013, 2017). PP holds that the brain uses unsupervised methods of learning to train its networks. The trained networks capture the essential features of reality. In a nutshell, the unsupervised methods of learning—such as what Hebbian theory offers—indicate that the intrinsic strength or weight of a given synaptic connection is proportional to the repeated stimulation of the post-synaptic neurone by the presynaptic neurone. Connections that are not exposed to frequent stimulations of their respective neurones would be gradually extinguished. There are computational strategies—e.g., Helmholtz’ machine and the wake-sleep algorithms—that
123
M. D. Beni
adequately formulate the brain’s strategies for unsupervised learning (Dayan et al. 1995). Thus PP can explain how the brain captures the causal structure of reality by forming prior expectations and updating them in accordance with the upward coming torrents of sensory inputs. According to PP, the brain generates predictive models of the world and tries to minimise the predictive error by using ‘active inferences’. PP regards the brain as a statistical engine that allows for self-generation of Bayesian models. The brain invokes Bayesian strategies so as to compare its models with their targets in the external world and modify them by reducing their predictive error. There is a dynamical interplay between top-down predictions of the brain’s models on the one hand, and bottom-up torrents of the actual sensory inputs (i.e., predictive errors) on the other. Karl Friston and friends introduced the free-energy formulation of PP. The free energy principle itself is defined as an information-theoretic measure that bounds the surprise on sampling some data, given a generative model (Friston 2010, 1). A ‘generative model’ is a probabilistic model of the dependencies between causes and consequences (data), from which samples can be generated. The generative models are conceived in terms of likelihoods and priors, where likelihoods are the probability of sensory data (given their causes), and a prior is the a priori probability of those causes. Perception, consists of the inversion of the likelihood model to access the posterior probability of the causes, given sensory data which amounts to mapping from sensations to causes (Friston 2010, 3). The theory provides the measure of the discrepancy between the causal structure of the world and the brain’s (or the organism’s) representation of that structure. Active and affective capacities of the brain emerge as a result of the organism’s success at minimising its free energy. Biological systems try to retain their composition in the face of the changing environment by regulating their internal environments to maintain their states within bounds (ibid). In this fashion, the organisms seek to reduce the element of surprise. The surprise is defined as the negative log-probability of an outcome, where an improbable outcome, e.g., being outside water for a fish, is surprising. So, surprise leaves a negative effect on the life of the organism, and by minimising the free energy, the biological system succeeds in minimising surprise. Friston also argued that ‘‘Agents can suppress free energy by changing the two things it depends on: they can change sensory input by acting on the world or they can change their recognition density by changing their internal states’’ (Friston 2010, 3 emphasis added). To make a long story short, PP is capable of explaining how the general experience of human agents can be codified in terms of probability distributions over possible observations. PP explains how the brain succeeds in representing the causal structure of the world, at least to the extent that is necessary for an organism’s survival in the world. Therefore, at a foundational level, PP helps us to dissolve the latching problem in a naturalistically plausible manner. The solution does not rely on Tarskian or model-theoretic formal semantics. Nor does it invoke semantic information. This assertion can make the notion of truth itself (traditionally a semantic notion) redundant. But we aimed to dispense with semantics and the involved notion of truth. Therefore it is not a flaw of this proposal that it does not
123
Syntactical Informational Structural Realism
make the solution to the latching problem dependent upon the notion of truth. It is rather its feature that it does not rely on semantics. More importantly, the unified entropy-based informational framework can embed the information-theoretic measure of free energy adequately enough. Friston’s free energy principle proposes to account for the relationship between the brain and the environment on the basis of Shannon entropy (as a probability distribution over all possible states that could be chosen by the organism as its environment). Therefore PP can be assimilated into the unified entropy-based framework. Let me elaborate. As we have already seen, Adriaans pointed out that Gibbs’ account of entropy defines ‘free energy’ in terms of the distance between the entropy of the actual system and maximal entropy (Adriaans 2010, 44). And it is possible to assume that Friston’s free energy is connected to Gibbs free energy. Shannon entropy was inspired by thermodynamic entropy, and as Adriaans (2010) pointed out, Gibbs entropy is compatible with Shannon’s conception of information. Since Friston’s information-theoretic notion of free energy is compatible with Shannon entropy, it could be claimed that Friston’s free energy is also connected with Gibbs’ notion of entropy, which is a thermodynamical concept. Of course, Gibbs’ free energy is defined in the context of classical thermodynamics whereas Friston’s concept of free energy is not (primarily) a thermodynamic concept. But although the connection between these two notions of free energy is not total, it can be asserted that Gibbs’ free energy and Friston’s free energy are to some extent compatible. But even without assuming that the two involved notions of free energy are identical, it could be granted that the unified entropy-based framework of information can assimilate Friston’s free energy principle and use it to latch onto the world. It remains true that free energy principle says nothing about the meanings that the agents draw from the world. But from the perspective of SISR, being independent of meaning and semantical factors is a feature of PP, which makes it an ideal candidate for underpinning the pragmatic links between theories (which are produced in the brain of scientists) and reality. In the next section, I shall put further emphasis on the pragmatic leaning of PPbased account of the brain-world relationship. I shall also explain how the pragmatic approach that we developed in this paper is consistent with the constructionist core of Floridi’s ISR. 6.2 Action-oriented-ness of Perception PP could be construed in two different ways. The first construal, which has been advocated by Jakob Hohwy, underlines the Helmholtzian origins of PP and highlights its inferentialist-internalist aspects (Hohwy 2013, 2014 among other places). Although Hohwy’s interpretation does not overlook the role of active inferences, it assumes that the active inferences are still inferences and the involvement of the action does not change the inferential nature of the brain-world relationship. Hohwy’s construal is centred on the notions of inference and representation. However, it does not explicitly associate the brain-world relationship with Tarskian or model-theoretic formal semantics. Nor does it invokes semantic information. Therefore, even this inferentialist construal is not openly inconsistent
123
M. D. Beni
with the pragmatic leaning of SISR. There is, however, also another construal which meshes nicely with my pragmatic stance. The alternative construal has been offered by Clark (2015, 2016a, b). Clark’s construal relies on enactivism (and it includes elements from the extended mind and the embodied cognition theses).7 This enactivist construal underlines the intertwinement of the representational mechanisms of perception with sensorimotor and motor control procedures. It seeks to obliterate the inferential veil, which was the central factor of Hohwy’s inferentialism, and it highlights the role of the situated agent and its evolutionary-ecological goals in the formation of cognition. As Clark’s (2013, 13) pointed out, since the organisms live and forage in a changing and challenging world, they have to learn to hence ‘‘expect’’ to deploy quite complex ‘‘itinerant’’ strategies to stay within our species-specific window of viability. Change, motion, and search are essential features of the life of creatures living in a world in which the resources are unevenly spread, and new threats and opportunities continuously arise. And the organisms are usually acquired the necessary skills to deal with the situation. Recently, Bruineberg et al. (2016) endeavoured to develop the enactivist reading of PP. Clark’s view held that the predictive brain aims at reducing the complexities of neural processing by means of contriving action routines in a world built of affordances-opportunities for action and intervention (Clark 2016a). Highlighting the point that self-maintenance and self-organisation are the defining features of the self-producing systems, Bruineberg et al. (2016, 4 ff) suggested that there might be an intimate relation between free energy minimisation and the biological features of the living systems (or perhaps the definition of the life itself). This allows for construing the dynamical interplay between the organism and the environment in terms of John Dewey’s pragmatism and James Gibson’s ecological psychology (6). The free energy principle applies to the organic systems. Organic systems could keep the amount of their free energy in check by making their internal dynamics conform to the environmental dynamics (perception) or by making the environment conform to the internal dynamics (action). To adapt to the environment, the organism changes its internal dynamics. On the other hand, to avoid finding itself in the biologically unbearable state, the organism seeks to change its sensory states through acting in the world. Hence a dynamical interaction between the organism and its environment. The organism is open to the environmental affordances (in the Gibsonian sense) and promotes its action and cognition by what the environment bestows upon it. This ecological reading indicates that instead of representing the causal structure of the world inside of a brain decoupled from the environment, PP aims to account for the embodied agent’s skills and affordance-related states of action-readiness which are coupled with the environment. Free energy framework explains the animals’ tendency ‘‘to act by the relevant affordances in its situation, so 7
Enactivism and embodied mind thesis define cognition as embodied action and underline the role agents as dynamical systems enacting in the world. Enactivism could be traced back to the continental traditions that underscores the role of body in the cognition (see Varela et al. 1991). Extended mind thesis is stemmed out of Clark and Chalmers’ active externalism and their emphasis on the contribution of the brain and environment in forging a coupled cognitive system (Clark and Chalmers 1998).
123
Syntactical Informational Structural Realism
as to reduce this tension, which results in its improving its grip on its environment’’ (10). 6.3 On the Pragmatic Nature of the Solution PP underpins the scientifically-informed pragmatics-based reply to the latching problem. Both salient interpretations of PP (offered by Hohwy and Clark) are compatible with a pragmatic account of the relationship between informational structures and the world (although the enactivist construal offered by Clark and Bruineberg et al. meshes more nicely with the pragmatic solution to the latching problem). In this subsection, I clarify the pragmatic nature of my solution. The latching problem is usually understood as a semantical problem. It may be assumed that even if a pragmatic solution can deal with the latching problem, it somewhat implements a semantical solution by definition. The vagueness of the border between semantics and pragmatics and the existence of a connection between the two domains may reinforce this suspicion. Although this suspicion may be justified, it is not the aim of this paper to obliterate the difference between the domains of semantics and pragmatics on the basis of the vagueness of their borders. While I would be happy to accept that the argumentation of the paper broadens the borders of the more traditional forms of semantics, I do not want to go so far as to refuse the existence of a general distinction between the domains of semantics and pragmatics. Therefore, I assert that the solution that I proposed here is not semantical in the model-theoretic sense. Nor is it semantical in the sense of SSTI. While my solution to latching problem is compatible with Floridi’s constructionist approach to semantic information, it (i.e., mine) is more thoroughly pragmatic (and naturalistic) than what has been offered in Floridi’s theories. The similarity between action-oriented construal of PP and theories of classical pragmatists such as John Dewey has been spotted previously (Bruineberg et al. 2016; Engel et al. 2013). Although a number of classical pragmatists, such as Dewey, Peirce, and Charles Morris were among the main contributors to the scientific field of pragmatics, pragmatism and pragmatics are not the same. Pragmatics is a discipline of semiotics, and it is concerned with the role of context in the formation of (the speaker) meaning. Pragmatics is an alternative to syntax (concerned with the form) and semantics (concerned with reference and designation). In this paper, I used ‘pragmatic’ with an eye to this explicated sense of pragmatics. Among others, Feigl (1950, 49) explicated this sense of pragmatics in terms of ‘‘the psycho-bio-sociology of cognitive behaviour’’ of organisms in general and humans in particular. This seems to be the same sense that is at issue in the works of Carnap (1955) and some other philosophers and linguistics (Morris 1938; Oller 1972; Uebel 2013). That being said, I have to add that the pragmatic solution that I proposed in this paper is compatible with the attempts at broadening the borders of the existing forms of semantics. More specifically, it is compatible with Floridi’s constructionist view. Floridi’s constructionist conception of semantic information, too, takes into consideration the role of affordances in underlying the open and dynamical interaction between the agent and the world. He asserted that semantic information
123
M. D. Beni
is neither only in the environment nor only in the mind but arises from the interactions of specific agents within their environment (Floridi 2011b, 295). Floridi’s account of the human-world relationship is agent-based, and it emphasises the point that agent’s openness to what the world may offer (as the niches of affordance) plays a role in forming the agent-world relationship (Floridi 2009, 176). The same conception lies at the roots of Floridi’s (2014) attempt at explaining how the agent elaborates and understands information. This problem is identifiable as the notorious symbol grounding problem (SGP). This is the problem of explaining how agents can elaborate their own semantics (autonomously) for symbols and signals that are the currency of their interaction with their environments and other agents, without relying on innatism or externalism8 (Harnad 1990; Taddeo and Floridi 2005). Floridi’s own ‘‘praxical’’ solution to SGP was presented in terms of a constructionist and action-based form of semantics (Taddeo and Floridi 2007). The solution is centred on the assumption of the existence of interactions between the agent and the environment, where the environment is the source of data used by the agent to constrain the affordances to create semantic information of the environment. The solution indicates that semantic information is the outcome of the agent’s active and constructive interpretation of the system that is the referent/source of the relevant data, not of its passive representation. Floridi’s view lines up with the cognitive-pragmatic solution to the latching problem. As my brief hints at Floridi’s works indicate, SISR can be quite close to the forms of informational realism that are advocated by Floridi. The main difference is that on some occasions (including his two papers on ISR) Floridi drew on the resources of computer science to articulate his conception of semantics, whereas, I inform my solution by drawing on the resources of empirical psychology and computational neuroscience. As my references in this section indicate, on some occasions, Floridi, too, added some biological realist tendency to his engagement with informational realism. That is to say, on some occasions, Floridi’s informational realism tends towards an ecological approach. So far as I can see, this inclination has not been integrated into Floridi’s version of ISR. But I assume that SISR has a lot in common with a possible extension of Floridi’s pragmatic and biological realist engagement with information theory.
7 Conclusion In this paper, I outlined a syntactical alternative to Floridi’s ISR. SISR replaces the semantic component of ISR with a pragmatic solution to the latching problem. I argued that the pragmatic solution to the latching problem is not only possible but also naturalistically plausible. I fleshed out the argument on the basis of the brain’s predictive coding theory. The information-processing of the cognitive systems underpins the formation and verification of scientific theories. Informational structures of the scientific theories, which could be regimented by the unified 8
This indicates that the data should not be either preinstalled in the cognitive system or uploaded from an external source.
123
Syntactical Informational Structural Realism
entropy-based informational framework, latch onto reality on the basis of the predictive coding capacity of the brain. Thus we can account for the connection between the unified entropy-based informational framework (which regiments the informational structure of theories) and the world on the basis of the brain’s capacity for decreasing the discrepancy between its models and reality. I finally argued that despite some possible disagreements over the issues of naturalism and semantics, ISR and SISR could form an alliance against the more orthodox, representationalist branches of SR.
References Adriaans, P. (2010). A critical analysis of Floridi’s theory of semantic information. Knowledge, Technology & Policy, 23(1–2), 41–56. https://doi.org/10.1007/s12130-010-9097-5. Beni, M. D. (2016). Epistemic informational structural realism. Minds and Machines, 26(4), 323–339. https://doi.org/10.1007/s11023-016-9403-4. Bruineberg, J., Kiverstein, J., & Rietveld, E. (2016). The anticipating brain is not a scientist: The freeenergy principle from an ecological-enactive perspective. Synthese. https://doi.org/10.1007/s11229016-1239-1. Carnap, R. (1937). The logical syntax of language. Edited by A. Smeaton (trans.). London: Kegan Paul Trench, Trubner & Co. Carnap, R. (1942). Introduction to semantics: And formalization of logic. Cambridge, MA: Harvard University Press. Carnap, R. (1955). Meaning and synonymy in natural languages. Philosophical Studies. https://doi.org/ 10.1007/BF02330951. Clark, A. (2012). Dreaming the whole cat: Generative models, predictive processing, and the enactivist conception of perceptual experience. Mind, 121(483), 753–771. https://doi.org/10.1093/mind/ fzs106. Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36(3), 181–204. https://doi.org/10.1017/s0140525x12000477. Clark, A. (2015). Radical predictive processing. The Southern Journal of Philosophy, 53(S1), 3–27. https://doi.org/10.1111/sjp.12120. Clark, A. (2016a). Surfing uncertainty. Oxford: Oxford University Press. https://doi.org/10.1093/acprof: oso/9780190217013.001.0001. Clark, A. (2016b). Busting out: Predictive brains, embodied minds, and the puzzle of the evidentiary veil. Nouˆs,. https://doi.org/10.1111/nous.12140. Clark, A., & Chalmers, D. (1998). The extended mind. Analysis, 58(1), 7–19. https://doi.org/10.1093/ analys/58.1.7. da Costa, Newton C. A., Bueno, O., & French, S. (1998). The logic of pragmatic truth. Journal of Philosophical Logic, 27(6), 603–620. https://doi.org/10.1023/a:1004304228785. da Costa, Newton C. A., & French, S. (2003). Science and partial truth. Oxford: Oxford University Press. https://doi.org/10.1093/019515651x.001.0001. Dayan, P., Hinton, G. E., Neal, R. M., & Zemel, R. S. (1995). The Helmholtz machine. Neural Computation, 7(5), 889–904. Elias, P., Carnap, R., & Bar-Hillel, Y. (1954). An outline of a theory of semantic information. The Journal of Symbolic Logic, 19(3), 230. https://doi.org/10.2307/2268645. Engel, A. K., Maye, A., Kurthen, M., & Ko¨nig, P. (2013). Where’s the action? The pragmatic turn in cognitive science. Trends in Cognitive Sciences, 17(5), 202–209. https://doi.org/10.1016/j.tics.2013. 03.006. Feigl, H. (1950). Existential hypotheses. Realistic versus phenomenalistic interpretations. Philosophy of Science, 17(1), 35–62. Floridi, L. (2004). Outline of a theory of strongly semantic information. Minds and Machines, 14(2), 197–221. https://doi.org/10.1023/b:mind.0000021684.50925.c9.
123
M. D. Beni Floridi, L. (2005). Is semantic information meaningful data? Philosophy and Phenomenological Research, 70(2), 351–370. https://doi.org/10.1111/j.1933-1592.2005.tb00531.x. Floridi, L. (2008). A defence of informational structural realism. Synthese. https://doi.org/10.1007/ s11229-007-9163-z. Floridi, L. (2009). Against digital ontology. Synthese, 168(1), 151–178. https://doi.org/10.1007/s11229008-9334-6. Floridi, L. (2011a). Semantic information and the correctness theory of truth. Erkenntnis, 74(2), 147–175. https://doi.org/10.1007/s10670-010-9249-8. Floridi, L. (2011b). A defence of constructionism: Philosophy as conceptual engineering. Metaphilosophy, 42(3), 282–304. https://doi.org/10.1111/j.1467-9973.2011.01693.x. Floridi, L. (2014). Perception and testimony as data providers. Logique et Analyse, 57(226), 3421–3438. Floridi, L. (2016). What a maker’s knowledge could be. Synthese. https://doi.org/10.1007/s11229-0161232-8. Floridi, L. (2017). A plea for non-naturalism as constructionism. Minds and Machines. https://doi.org/10. 1007/s11023-017-9422-9. French, S. (2014). The structure of the world: Metaphysics and representation. Oxford: Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199684847.001.0001. French, S. (2015). (Structural) realism and its representational vehicles. Synthese. https://doi.org/10.1007/ s11229-015-0879-x. French, S., & Ladyman, J. (1999). Reinflating the semantic approach. International Studies in the Philosophy of Science, 13(2), 103–121. https://doi.org/10.1080/02698599908573612. French, S., & Ladyman, J. (2003). Remodelling structural realism: Quantum physics and the metaphysics of structure. Synthese, 136(1), 31–56. https://doi.org/10.1023/A:1024156116636. French, S., & Ladyman, J. (2011). In defence of ontic structural realism. Scientific Structuralism, 281, 25–42. https://doi.org/10.1007/978-90-481-9597-8_2. Friston, K. J. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138. https://doi.org/10.1038/nrn2787. Friston, K. J. (2012). Predictive coding, precision and synchrony. Cognitive Neuroscience, 3(3–4), 238–239. https://doi.org/10.1080/17588928.2012.691277. Friston, K. J., & Price, C. J. (2001). Generative models, brain function and neuroimaging. Scandinavian Journal of Psychology, 42(3), 167–177. Friston, K. J., & Stephan, K. E. (2007). Free-energy and the brain. Synthese, 159(3), 417–458. https://doi. org/10.1007/s11229-007-9237-y. Frith, C. D. (2007). Making up the mind : How the brain creates our mental world. Malden: Blackwell Pub. Harnad, S. (1990). The symbol grounding problem. Physica D: Nonlinear Phenomena, 42(1–3), 335–346. https://doi.org/10.1016/0167-2789(90)90087-6. Hohwy, J. (2013). The predictive mind. Oxford: Oxford University Press. https://doi.org/10.1093/acprof: oso/9780199682737.001.0001. Hohwy, J. (2014). The self-evidencing brain. Nouˆs, 50(2), 259–285. https://doi.org/10.1111/nous.12062. Hohwy, J. (2017). How to entrain your evil demon. In T. Metzinger & W. Wiese (Eds.), Philosophy and predictive processing. MIND Group: Frankfurt, Frankfurt am Main. https://doi.org/10.15502/ 9783958573048. Morris, C. (1938). Foundations of the theory of signs. Chicago, Ill: University of Chicago Press. Oller, J. W. (1972). On the relation between syntax, semantics, and pragmatics. Linguistics, 10(83), 43–55. https://doi.org/10.1515/ling.1972.10.83.43. Rao, R. P., & Ballard, D. H. (1999). Predictive coding in the visual cortex: A functional interpretation of some extra-classical receptive-field effects. Nature Neuroscience, 2(1), 79–87. https://doi.org/10. 1038/4580. Shannon, C. E., & Weaver, W. (1949). The mathematical theory of communication. Urbana, IL: University of Illinois Press. Srinivasan, M. V., Laughlin, S. B., & Dubs, A. (1982). Predictive coding: A fresh view of inhibition in the retina. Proceedings of the Royal Society of London B: Biological Sciences, 216(1205), 427–459. Stegmu¨ller, W. (1979). The structuralist view of theories. Berlin, Heidelberg: Springer. https://doi.org/10. 1007/978-3-642-95360-6. Suppe, F. (1998). Understanding scientific theories: An assessment of developments, 1969–1998. In Philosophy of Science Biennial Meetings of the Philosophy of Science Association. Part II: Symposia Papers (Vol. 67, pp. 102–115). http://www.jstor.org/stable/188661.
123
Syntactical Informational Structural Realism Suppes, P. (1962). Models of data. In Logic, methodology, and philosophy of science: Proceedings of 1960 International Congress (pp. 252–261). Stanford U: Stanford University Press. Suppes, P. (1967). What is a scientific theory?’’ In S. Morgenbesser (Ed.), Philosophy of science today (pp. 55–67). New York: Basic Books. https://www.google.com/_/chrome/newtab?espv=2&ie=UTF8. Taddeo, M., & Floridi, L. (2005). Solving the symbol grounding problem: A critical review of fifteen years of research. Journal of Experimental & Theoretical Artificial Intelligence, 17(4), 419–445. Taddeo, M., & Floridi, L. (2007). A praxical solution of the symbol grounding problem. Minds and Machines, 17(4), 369–389. https://doi.org/10.1007/s11023-007-9081-3. Tarski, A. (1944). The semantic conception of truth: And the foundations of semantics. Philosophy and Phenomenological Research, 4(3), 341. https://doi.org/10.2307/2102968. Uebel, T. (2013). Pragmatics in Carnap and Morris and the bipartite metatheory conception. Erkenntnis, 78(3), 523–546. https://doi.org/10.1007/s10670-011-9352-5. van Fraassen, B. C. (1980). The scientific image. Oxford: Oxford University Press. https://doi.org/10. 1093/0198244274.001.0001. Varela, F. J., Thompson, E., & Rosch, E. (1991). The embodied mind : Cognitive science and human experience. Cambridge: MIT Press. Worrall, J. (1989). Structural realism: The best of both worlds? Dialectica, 43(1–2), 99–124. https://doi. org/10.1111/j.1746-8361.1989.tb00933.x.
123