Acta Anal https://doi.org/10.1007/s12136-017-0333-4

The Downward Path to Epistemic Informational Structural Realism Majid Davoody Beni 1

Received: 27 January 2017 / Accepted: 2 October 2017 # Springer Science+Business Media B.V. 2017

1 Introduction1 Structural realism (SR) is a dominant view of the contemporary philosophy of science. It holds that empirical success and explanatory power of our best scientific theories are due to the relation of scientific theories to the Bmindindependent^ world (French 2014, p. 1). To the contrary, antirealists argue that scientific theories could not (nor do aim to) provide precise descriptions of mindindependent reality. The advocates of SR need to justify plausibility of their metaphysical claims. Some structural realists (e.g. Ladyman and Ross 2007) offered to justify plausibility of their metaphysics on naturalistic grounds. Accordingly, they suggested that we have to tailor our metaphysics to what our best sciences offer, instead of domesticating our sciences to traditional metaphysics. Ladyman and Ross (2007) also suggested that it is possible to account for the theories-world relationship by invoking information patterns. If the informational structures could be grounded in mind-independent reality, then real informational patterns that correspond to mind-independent structure of the world would be distinguishable from mere nominal ones. Luciano Floridi (2008, 2009, 2011) endeavoured to present an alternative form of informational SR (ISR for short). There is a significant difference between Ladyman et al.’s informational approach to SR and Floridi’s ISR. Floridi’s informational philosophy is centred on constructionism which holds that knowledge is mainly a form of constructive activity. According to this view, it is not the aim of knowledge to represent the

1

I greatly benefited from the comments and suggestions of the anonymous referees of this journal. One of them was particularly helpful with numerous insightful comments. It was not possible to promote the arguments of the paper without these suggestions. The debts are gratefully acknowledged.

* Majid Davoody Beni [email protected]

1

Department of Management, Science and Technology, Amirkabir University of Technology, No. 424, Hafez Street, PO Box 3313-15875, Tehran, Iran

M. D. Beni

phenomena passively. Knowledge, as a constructive activity, aims to create approximately correct informational models (i.e. semantic artefacts) of phenomena through a process of proactive interaction (Floridi 2017, p. 1). Accordingly, Floridi’s ISR forsakes the goal of representationalism and naturalism, and it leaves the orthodox metaphysics of OSR. While ISR offers to reconcile epistemic SR (ESR) to ontic SR (OSR), its break with the more familiar forms of SR seems to be radical. But could ISR be developed along the lines of ontic structuralism? Ladyman et al.’s ISR is compatible with the more orthodox forms of OSR. Their ISR 2 is committed to naturalism and representationalism, and it is presented as a version of (non-eliminativist) ontic structuralism. But the present paper aims to demonstrate that—in the absence of a viable strategy for naturalising information— Ladyman et al.’s informational version of OSR fails to justify its inflated metaphysics. Instead, this paper takes a downward path to a third option, i.e. epistemic ISR (EISR for short). EISR is a weaker stance. It is more modest than the outcome of Floridi’s unification between the ontic and the epistemic forms of SR. EISR leans towards ESR (see Beni 2016). By speaking of a downward path, 3 I allude to a strategy which starts from OSR. But because OSR’s metaphysical claims could not be substantiated plausibly, we have to replace ISR with EISR. EISR retains the naturalistic and representationalist basis of SR but does not make ontological claims about the mind-independent world. Since the goal of naturalisation of information has not been fulfilled satisfactorily, we have to dispense with the ontic component of SR and make room for the considerations of humility. The paper consists of three parts. In the first part, I outline a broad-brush sketch of SR. In the second part, I canvass two already existing naturalising strategies. These are Taddeo and Floridi’s (2007) praxical strategy and Ladymand and Ross (2007) attempt at distinguishing real patterns from mere patterns. I argue that neither of these two strategies could naturalise information satisfactorily.4 I also explain how EISR dissents from the ontic version of ISR. In the third part, I will assess the capacity of biosemiotics for naturalising semantic information in a non-derivative way, i.e. without mandating an interpreter or informee. I will argue that despite its realistic implication, even this last resort solution is not completely successful. I will conclude the paper by pointing out that information has not been naturalised satisfactorily enough to support the ontic claims of the OSR-theorist. It will be possible to choose Floridi’s constructionist approach. But given its constructionist undertone, Floridi’s ISR breaks with the more familiar forms of SR. On the other hand, EISR is compatible with Kantianism and Floridi’s constructionism but it does not cut its ties with the standard version of ESR and the ideals of naturalism and representationalism. 2

To underline the difference between Floridi’s and Ladyman et al.’s respective views, in this paper, I keep the abbreviation (i.e. ISR) for Floridi’s version and call Ladyman et al.’s version informational OSR. 3 The notion of Bdownward path^ is adopted from Stathis Psillos (2001) who marked a distinction between upward path and downward path to SR. The upward path begins from an empiricist position and negotiates its way to a structural realist stance, whereas the downward path begins from a strongly metaphysical stance and winds up with a modified and modest version of SR. 4 I am told that at least according to some non-eliminativist OSR-theorists (such as Ladyman et al.) what is really needed is a naturalisation of representation, say, instead of a naturalisation of information. Long (2014) has articulated the distinction clearly. While I understand the distinction, I think wrestling with the problem at the level of naturalisation of information, if possible at all, could result in a more straightforward solution to the problem of substantiating the ontological claims of ISR-theorists. This does not mean that the distinction itself is not tenable.

The Downward Path to Epistemic Informational Structural Realism

2 Structural Realism, and Its Varieties SR is a popular theory of the contemporary philosophy of science. It has been developed into different branches, e.g. epistemic, ontic, eliminativist, non-eliminativist, and ISR. Structural realists of various stripes generally accept that our scientific theories provide approximately precise descriptions of (the underlying structure of) reality. ESR is the extension of Kantian agnosticism with regard to foundational ontology, and its main claim is about what we could fundamentally know about the world’s structure. ESR emerged as a reaction to the pessimistic meta-induction, which holds that success of the present scientific theories is no guarantee of their truth, because even the past false theories had been presumed to be true. This provides the basis of the antirealist’s argument. According to this argument, given the historical discontinuity of the scientific theories, we may doubt truth-conduciveness of the present theories. ESRtheorists reacted by pointing out that the continuity could be found at the level of form of the theories, not their content (Worrall 1989). ESR provides best of the both worlds of realism and antirealism. The view agrees with antirealism in holding that we could not make ontological commitments to underlying reality, but it agrees with realism in accepting that we have to make epistemological commitments to the underlying structure of reality. OSR goes beyond ESR’s agnosticism. OSR holds that all that there is, is structure. OSR has been defended on the basis of the evidence from the contemporary physics. Modern physics (i.e. quantum mechanics, quantum field theory) eschews concept of the individual object. Accordingly, a realist may want to change her loyalties and make ontological commitments to the commonalities that lie beneath the non-individual (and hence structural) objects (French and Ladyman 2011). Then again, there is a division of opinion on retaining a thin notion of individual objects in the ontology or eliminating them completely (Esfeld and Lam 2008; Muller 2011). SR has also been extended into an informational branch. Ladyman et al. used information theory to lay the groundwork of representational and metaphysical aspects of OSR (Ladyman and Ross 2007). They articulated naturalised metaphysics of OSR. Traditional metaphysics holds that it is possible to proceed a priori in investigation of the metaphysical matters upon which science does not bear. But naturalised metaphysics denies that there is any good reason for thinking that a priori metaphysical knowledge is possible. Naturalised metaphysicians argue that priori metaphysics has achieved nothing remotely comparable with the breakthroughs in science and mathematics, and in this sense they favour contingency over a-prioricity (Ladyman and Ross 2007). They also argue that minimal metaphysics consists in an antifoundationalist version of unification of sciences (as what constitutes the metaphysics of OSR). To achieve this goal, Ladyman and Ross (2007, chapter 4) endeavoured to account for the requisite intertheoretical relations in terms of information channels. Ladyman et al.’s informational structuralism is a form of non-eliminativist OSR (NOSR). Almost simultaneously, Floridi constructed his version of ISR. Floridi developed ISR through a bipartite project. In the parsconstruens, he defended ISR in the following terms: Explanatorily, instrumentally, and predictively successful models (especially, but not only, those propounded by scientific theories) at a given LoA [i.e. Level of

M. D. Beni

Abstraction] can be, in the best circumstances, increasingly informative about the relations that obtain between the (possibly sub-observable) informational objects that constitute the system under investigation (through the observable phenomena). (2008, pp. 240–241) This stance is stemmed out of Floridi’s refutation of the digital ontology in the parsdestruens of the project (Floridi 2009, also 2011, chapter 14), wherein he argued that conceiving of reality in digital vs. analogue terms is a categorical mistake. Let me elaborate. According to Floridi, digital and analogue are only the modes of Bpresentation^ of Being, as being experienced or conceptualised by an epistemic agent who interacts with reality, at a given level of abstraction (LoA). Floridi primarily presented LoAs as frameworks of epistemological inquiry. In its technical sense (derived from computer engineering), a LoA is a set of typed variables. It consists in an interface that establishes the type and scope of data that could be used as a source for generation of information (Floridi 2011). A typed variable is Ba uniquely-named conceptual entity (the variable) and a set, called its type, consisting of all the values that the entity may take^ (Floridi 2011, p. 48). Each adopted LoA delimits the border of inquiry in a fundamental way. This means that from the outset, they delimit range of the questions that (a) can be meaningfully asked and (b) are answerable in principle (Floridi 2009, p. 166). Arguably, this may result in what Floridi (2011) called Bontological levelism^ (the same phenomenon has been called Bscale relativity of ontology^ by Ladyman and Ross (2007 chapter 4)). Ontological levelism is consistent with the constructionist spirit of Floridi’s enterprise. According to Floridi: B[r]eality is experienced, conceptualised and known as digital or analogue depending on the level of abstraction (LoA) assumed by the epistemic agent when interacting with it. Digital and analogue are features of the LoA modelling the system, not of the modelled system in itself^ (Floridi 2009, p. 160; 2011, p. 325). However, metaphysical demands of OSR are stronger than what can be satisfied by Floridi’s constructionist approach to realism. This is because OSR-theorists agree that, Bgenerally speaking the scientific realist accepts that there is a mindindependent reality ‘out there’^ (French 2014, p. 1). Being agreeable to this common wisdom, OSR-theorists also consent that what is responsible for the empirical success and explanatory power of our best scientific theories is their relationship the mindindependent world (French 2014, p. 1). This form of representationalism is at odds with Kantian origins and constructionist tendency of Floridi’s informational ontology and his emphasis on the role of the LoAs in determining the question of ontology. Floridi seems to be quite conscious of the kerbing connotation of implementing ontological claims within an epistemic framework; Bontological commitments are initially negotiated through the choice and shaping of LoAs, which therefore cannot presuppose a metaphysical omniscience^ (Floridi 2011, p. 72). Ladyman et al.’s version of informational OSR seems to be constrained by the same limitations that the informational ontology sets on Floridi’s approach, too. But representationalist and naturalist commitments of Ladyman et al.’s version of OSR are made too fundamentally to be dismissed easily. The point that has to be taken into consideration is that Floridi’s notion of ontological levelism has a counterpart within the context of the Ladyman et al.’s informational approach to OSR. I am referring to the notion of the scale relativity of ontology.

The Downward Path to Epistemic Informational Structural Realism

The scale relativity (when applied to either epistemology or ontology) is the idea that Bwhich terms of description and principles of individuation we use to track the world vary with the scale at which the world is measured^ (Ladyman and Ross 2007, p. 199). Here, the Bscale^ fills the shoes of Floridi’s notion of BLoA.^ It follows that the scale relativity of ontology (almost equal to ontological levelism) could be a hazardous hypothesis from a metaphysical realist point of view. It holds that what exists mind-independently has to be relativized to scales at which nature is measurable. One may want to insist that the scales are real and mind-independent. But then again, saying of certain objects that they only exist at certain scales (or levels) can easily slide into a tendentious way of saying that the objects in question are not real at all (Ladyman and Ross 2007). The scale relativity of ontology has to be constrained. Otherwise, the possibility of distortion of the scientific representations by the influence of a specific scale at which the world is measured could not be suppressed. This endangers the assumption of the faithful correspondence between scientific theories and the scale-independent reality. One may want to appeal to criteria of informativeness, coherence, elegance, simplicity, explanatory power, consistency with the data, and predictive power in order to deal with the problem of scale relativity. However, these criteria may be understood as pragmatic, heuristic, or at any rate methodological and non-epistemic factors. This may indicate that instead of connecting the theory to reality with a steady and converging pace, these methodological factors pull in opposite directions and tear the realist framework apart. The criteria of informativeness and simplicity, for example, do not necessarily work together in order to choose the same theory. A similar worry has been raised by Bueno (2010), who submitted that the most basic ingredients of the ISR, i.e. informativeness and truth, might oppose one another. For, a literally false theory, such as Newtonian mechanics, could be very informative, while a true statement, such as a tautology of classical logic can be quite uninformative (Bueno 2010, p. 367). Moreover, if we grant that the scales have a significant bearing on the relationship between theories and reality, we cannot ascertain how loyal the theories are to the scale-independent reality that they are supposed to represent. As I remarked before, the problem of scale relativity or ontological relativism is not a problem for Floridi who embraced constructionism at the cost of naturalism and representation. However, this option is not available to Ladyman et al. who placed naturalism at the very centre of their metaphysical structuralism. Here, attempting the naturalisation of information may help the friend of informational OSR to elaborate a realistic account of the theories-world relationship. We shall survey the viability of this option.

3 Grounding the Informational Structures To account for the representational relation between the structure and reality, the advocates of informational OSR should be able to ground the informational structure of the theories in nature. We shall proceed to survey some strategies for dealing with the naturalisation problem. Floridi’s approach to informational realism does not need to be supplemented with a solution to the naturalisation problem (given his constructionist tendency). However, on a separate occasion, Floridi had attempted finding a reply to the symbol grounding problem (SGP). After assessing Floridi’s attempt at grounding information in the next section (Sect. 3.1), I will survey Ladyman et al.’s use of real patterns for the same purpose.

M. D. Beni

3.1 Floridi’s Attempt Floridi articulated the problem of naturalisation in terms of the SGP (Floridi 2011; Taddeo and Floridi 2005, 2007). This discussion has not been presented as directly relevant to ISR, and although Floridi’s engagement with SGP could affect ISR, it is primarily aimed at dealing with the issues of the nature of information and its semantic features. SGP could be construed as a special form of naturalising problem. Yet, SGP could be successfully solved, without any successful naturalisation of information. Below, I shall remark that Floridi’s solution to SGP, as being proposed in this section, seems to deal with the problem successfully enough within the context of AI and robotics without succeeding (or perhaps even aiming) at naturalising the information. Let me elaborate. Harnad spelt out the problem in following terms: B[h]ow can the semantic interpretation of a formal symbol system be made intrinsic to the system, rather than just parasitic on the meanings in our heads? How can the meanings of the meaningless symbol tokens, manipulated solely on the basis of their (arbitrary) shapes, be grounded in anything but other meaningless symbols?^ (Harnad 1990, p. 335). No forms of innatism or externalism are allowed, but the agent may have its own computational, perceptual, etc., resources to be able to ground its own symbols. These are called the zero semantical condition. Harnad’s own strategy for resolving the problem consisted in combining the strengths of the symbolic system and the connectionist model, but Taddeo and Floridi (2005) evaluated this as implausible. After reviewing several other strategies for dealing with the grounding problem, Taddeo and Floridi came to conclude that none of the strategies that has been thus far suggested is particularly successful in dealing with the SGP satisfactorily (Taddeo and Floridi 2005). Here, I avoid engaging in a detailed report of Taddeo and Floridi’s rebuttal. Suffice it to say that according to Floridi, Ball of these strategies either fail to address the SGP or circumvent it, by implicitly presupposing its solution and begging the question^ (Floridi 2011, p. 135). I proceed with focusing on Floridi’s solution to SGP. He called it praxical so as to stress the key role played by the interactions between the agents and their environment (Floridi 2011, p. 162). Floridi designed a new kind of artificial agent (AA) (actually a two-machine AA, AM2). He also developed a new theory of semantics, namely actionbased semantics, in order to account for the meaning relations that prevail between AM2s. Action-based semantics is supposed to outline the approaches that are involved in the process of coupling symbols to meaning (Floridi 2011, p. 163). AM2s implement action-based semantics, and this allows them to ground their symbols semantically. Meaning relations generated by the AAs are directly correlated with actions performed by the same AAs (Floridi 2011, p. 164). In this picture, the semantic process that is involved in the generation of the meaning of the symbols is supposed to be reducible to (or at least explicable in terms of) the immediate consequence of an AA’s performance (ibid). The internal states of AAs are, according to Floridi, excellent candidates for the role of non-semantic and yet semantic-inducing resources (ibid). As I remarked, AM2s have a two levelled architecture. The lower level, the M1, through which action is related to the world, interacts directly with the environment, while the M2, through which symbols are related to action, operates at the metalevel. There is interaction between the metalevel and the environment. The claim is that through this architecture, meaning could be ascribed to the symbols in a non-derivative

The Downward Path to Epistemic Informational Structural Realism

way. I have to impress the point that the notions of action and performance that are at issue in this theory do not elicit any teleological explanation, nor their relation to goaloriented behaviour is taken into account (Floridi 2011, p. 166). This is supposed to guarantee that Floridi’s solution for SGP is non-derivative. Floridi’s declaration (Floridi 2011, p. 167) notwithstanding, it is not quite true that the M1 interacts directly with the external environment, not unless one wants to understand LoA-mediated relation as direct relations. For Floridi held that the M1 interacts with the environment at a level of abstraction Bthat allows only a specific granularity of detection of its features^ (Floridi 2011, p. 168). So, the notion of the LoA figures in again, and it brings about the relevant notions of the observable (in the computer science sense of the observable) and the interpreted typed variable. Application of the notion may violate zero semantical condition. However, Floridi’s stress on AM2 being two machines may indicate that the AM2 itself is responsible for interpreting the typed variables. Reading the theory in this light may indicate that LoAs are not contrived by an external observer, but they belong to the machine itself. M2 operates at metalevel, and the target of its elaborations is the internal states of the M1, which interacts directly with reality. So, the emergence of semantics is the function of the performance and elaboration of the M2 (it is a function of the internal state of the M1, which is a function of the environmental information that is g (f (e)) (Floridi 2011, p. 171)). To indicate that this semantics is not private to AM2, Floridi attributed to AM2 some additional semantic abilities, such as communication with other similar machines, and capacity for originating a shared lexicon (Floridi 2011, p. 176), to the effect that meaning can be based upon interactions with other AM2s. (ibid, 180). Although Floridi’s solution might deal with SGP, it could not solve the problem of naturalisation of information in a way that is required for dealing with the problem of the scale relativity of ontology. The solution could not (and does not aim to) be developed to deal with the problem of grounding the informationally regimented structure of theories in nature. This is because the role of LoAs, as the epistemological-semantical frameworks, is still quite prominent in Floridi’s solution to SGP. Epistemic and semantic aspects are closely intertwined most of the time, and it seems unreasonable to claim that an epistemic-framework-mediated access to the reality is successful at capturing the mind-independent nature of the objective domain. Given the constructionist tendency of Floridi’s (2008, 2009) development of ISR, the failure of Floridi’s solution in naturalising information does not prevent Floridi’s ISR from being a foil to antirealism. But this solution to SGP could barely help establishing the ontological claims of the naturalistically launched defences of informational OSR. 3.2 Second Attempt Here, we focus on Ladyman et al.’s own attempt at grounding informational structures of the theories in nature. In the fourth chapter of Every Thing Must Go (2007), Ladyman et al. ventured to address the problem of scale relativity of ontology. The solution might pave the way to a straightforward reconstruction of OSR in accordance with the tenet of naturalism. According to Ladyman et al.’s development of naturalised metaphysics, OSR is the best synthesis between the views of scientific realists and antirealists on the topic of the relation between the success of scientific theories and the continuity of their underlying

M. D. Beni

structure through history (Ladyman and Ross 2007, chapter 2). Ladyman et al.’s account of the relation between the theories and reality is centred on the idea of informational (real) patterns. Their informational extension of OSR emerged as a realist and antireductionist reaction to Dennett’s idea of Breal patterns.^ According to Dennett, BA pattern exists in some data-is-real-if there is a description of the data that is more efficient than the bit map, whether or not anyone can concoct it^ (Dennett 1991, p. 34).The problem is that it is often possible to read different patterns into the same data, depending on the scale that one adopts to interpret the data. But it is not always possible to rely on the comparative superiority of a real pattern in predicting the value of some micro-scale data. This is because, in the absence of any other criterion (except the practical significance of the pattern), the observers’ preferred level of tolerance for noise in producing predictions may cause them to mark different patterns as real. Sliding back to instrumentalism is a consequence of tying the choice of real patterns to the observer’s idiosyncratic pragmatic interests. Ladyman et al. sought to overcome the problem by providing an empirically motivated story about how real patterns are distinguishable from patterns simpliciter. In a nutshell, their reply is that it is physics that constrains the number of patterns that are discernible by experience. To establish this reply, they endeavoured to weed out the Bintentional patterns^ that have been insinuated into pattern theory by Dennett (Ladyman and Ross 2007, p. 208). It was at this stage that the authors invoked the notion of the channels of information. In the light of Collier’s inspiring work in the 1990s, the authors claimed that being physically constrained is a condition on the requisite channels of information, and it is in this sense that physics determines what sort of channels could carry what sorts of information and what sorts cannot (ibid). Physics determines the lower bounds on the energy required for running effective computation. Thus, physics helps to eliminate the hypothetical computations that are physically impossible (Ladyman and Ross 2007, p. 208). In this vein, the authors provided an allegedly neat way for distinguishing the real patterns from the nominal ones. In other words, they provided a reliable way for removing the surplus structures, without violating their naturalistic principle. Moreover, the notion of informational channel does not make representational (or presentational) capacity of information dependent on semantic aspects of information. Let me elaborate. Floridi was an advocate of the strongly semantic theory of information (Floridi 2004, 2005). This means that he steered clear from both non-semantical mathematical theory of communication (MTC) (Shannon and Weaver 1949) and weakly semantic theory of information (Barwise and Seligman 1997; Dretske 1981; Elias et al. 1954). Some philosophers challenged the significance or even necessity of this semantical development (Adriaans 2010). But Ladyman et al.’s theory is free from the semantical complexities. They drew on John Collier’s philosophical attempt at reifying abstract Shannon-Weaver’s information (in terms of the MTC) as material or bounded information in the context of theories of physics, computability, and biology. Thus, bounded Shannon-Weaver information is the common currency that can be used for bringing unity to different domains, according to this view. The authors were clear that the relevant sense of information that is invoked in their ontology is Shannon-Weaver notion from the communication theory (Ladyman and Ross 2007, p. 220). MTC provides the quantitative measure of the amount of objective statistical uncertainty associated with the initial value of a message. To add some objectivity to their

The Downward Path to Epistemic Informational Structural Realism

repertoire, Ladyman et al. relied on notions from computer sciences such as Balgorithmic compressibility^ and the relevant notion of Blogical depth. 5^ Finding the suitable compression is the key to discerning the involved causal linkages. This is how the physically possible patterns (the real ones) are distinguishable from the hypothetically possible patterns. Once we decided that real patterns are distinguishable from nominal ones, it would be desirable to introduce the connective links that relate real patterns to the world. Ladyman et al. invoke Barwise and Seligman’s (1997) notion of flow of information to articulate the connective links. Ladyman et al.’s elaboration should appeal to those who (like Adriaans and Long) regard Floridi’s engagement with semantical aspects as superfluous. Below, I shall proceed to show that despite its advantages, Ladyman et al.’s approach could not substantiate the ontic component of OSR. As Ladyman et al. acknowledged, the real patterns are carrying information only about other real patterns (Ladyman and Ross 2007, p. 227). This could (and is intended to) mean that real patterns are the ultimate posits of ontological commitments. It may be assumed that ontology consists of real pattern all the way down. 6 However, to substantiate the ontic component of their SR, Ladyman et al. needed to demonstrate that the information channels are latched onto the mind-independent reality. Unless such further argument could be produced, Ladyman et al.’s statement of informational OSR would not deliver on the promise of accounting for the ontic component of their SR. Ladyman et al.’s account could not block the way of antirealism, given that antirealist could argue that the information channels connect the so-called real patterns with other models of data in physics, biology, computer science, etc., rather than with mind-independent reality. In response, Ladyman et al. could underline the unifying power of their metaphysical plan, by remarking that diverse disciplines are to be unified by the naturalistic metaphysics that contributes to the formation of a unified picture of science. Unity, however, is not ipso facto a grantor of the metaphysical realism. The antirealist does not need to demonstrate that no causal pathway exists between the objective world and models in our best scientific theories, in order to establish her pessimistic construal. She does not need to offer a positive proof of the impossibility of the existence of the source of information at the end of the channel. It is the realist who asserts that the source exists. Therefore, the burden of proof is on her shoulder. In the absence of a convincing argument from realist side, the antirealist could justify her pessimism by arguing that it is not the aim of science to make cumbersome ontological commitments to the existence of a mind-independent source of information. OSRtheorists may go further and use classical arguments (e.g. no miracle argument and inference to the best explanation) for their view. In reaction, the antirealist can also go further to challenge the validity of the conjured arguments (Ladyman et al. 1997; see van Fraassen 1980). Let us recap. The semantic preference that grounds Ladyman et al.’s ontological orientation is that reality is not a sum of concrete particulars, and our best access to reality is through the integrated patterns of experience. This is because individual 5

The logical depth is the normalised quantitative index of execution time required for generating a model of a real pattern by a near incompressible universal computer program, which is not itself computable as the output of a significantly more concise program. 6 Notice that Floridi’s metaphors of grey box and Russian doll (Floridi, 2008) suggest the same ontological hierarchy.

M. D. Beni

objects, events, and properties are devices used by observers to keep cognitive books on what science finds to be sufficiently stable or worth measuring over time, viz. some but not other patterns (Ladyman and Ross 2007, pp. 228–29). As persuasive as this argument may be in eliminating the individuals from ontology in favour of the structures (in structuralism vs. substantivalism debate), it does not provide any justification for traversing the gap between representations and reality. Given that the friend of OSR could not substantiate her inflated metaphysics naturalistically, one may decide to step back from OSR. EISR is an option that suggests itself to those of us who want to tread the downward route to a less inflated realist view, but at the same time do not want to make concession on an outright antirealist stance. In the next subsection, I shall allude to some reasons for taking the downward path. 3.3 On Modesty and Audacity I shall briefly explain why the metaphysical claims of OSR are inflated. I also articulate my foil to ontic ISR. Scientific realism could be articulated in different manners. According to Stathis Psillos, the more modest version of realism is centred on the claim about the existence of an independent and unobservable (by the senses) world which science tries to map. The more presumptuous claim is that Balthough this world is independent of human cognitive activity, science can nonetheless succeed in arriving at a more or less faithful representation of it, enabling us to know the truth (or at least some truth) about it^ (Psillos 2000, p. 707). An audacious version may hold that science represents everything that there is to be represented. SR is an extension of the traditional scientific realism. The more modest forms, such as ESR, try to comprise the best of both worlds of realism and antirealism. ESR holds that science succeeds in arriving at a more or less accurate (structural) description of the world. But ESR remains loyal to the spirit of Kantian modesty. It holds that we can know the structure of the world, but does not make claims about reality as such. OSR is far more presumptuous (ontologically). It holds that all that there is (independently of us), is structure. Ontic structural realists assume that since we may tailor our metaphysics to our epistemology—which is informed by scientific theories—we can assume that there are no quiddities or hidden natures, and our scientific theories represent all that there is. Metaphysics of OSR is inflated, in the sense that it cannot be justified by the arguments of its advocates. It could be shown that the arguments7 of the friends of OSR are not compelling (Morganti 2011). And there are positive arguments for humility. That is to say, it could be argued that dispositions, relations, and properties do not need to be exhausted in terms of the causal roles that they play in the constitution 7 OSR’s arguments consist of the argument from under-determination and the argument from the primacy of relations. I briefly allude to the former argument. According to the argument from underdetermination, modern physics implies that there is metaphysical underdetermination between two incompatible models of the world. One model (compatible with habituated intuitions) allows for the existence of individual objects. But quantum physics is not kind to individual objects and endorses an ontology in which individual objects are not fundamental. OSR proposes that we can get rid of the dilemma by dispensing with both (individual and non-individual) packages and making ontological commitments to the structures that underlie both models (French & Ladyman, 2003). Morganti (2011) denied the premises of the argument. He argued that it is not the case that postulating objects as fundamental entities leads to a problematic form of metaphysical underdetermination.

The Downward Path to Epistemic Informational Structural Realism

of the structures. The properties could possess functional roles and contribute to the mechanisms of a complicated causal nexus. But it is also conceivable that the properties possess intrinsic natures that are causally irrelevant to their functional aspects (see Whittle 2006). Under the circumstances, Kantian insights may inspire us to dissent from OSR and find an intermediary stance between the extremes of presumptuous metaphysics and radical empiricism. Given the context of the discussion of the paper, the intermediary stance is EISR. We do not even need to change the definition of Floridi’s ISR to articulate EISR. EISR holds that explanatorily, instrumentally, and predictively successful models at a given LoA can be, in the best circumstances, increasingly informative about the relations that obtain between the informational objects that constitute the system under investigation. If we lay emphasis on the informativeness (as an epistemological concept) and remain silent about ontic commitments of the theory, the definition will mesh nicely with ESR. There may be different ways for arriving at the intermediary stance of EISR. I recommend a downward path. Given that the metaphysical pretences of OSR are too grave to be substantiated, we may make concession on a less presumptuous and more modest version of ISR, which is EISR.

4 Biosemiotics and a Last Resort Solution Let us return to the SGP. It is a well-known fact that the SGP roots in John Searle’s Chinese Room thought experiment, which denies that AAs could possess nonderivative intentional states. More interestingly, as Searle himself remarked (Searle 1980), only biological systems have genuine intentionality. The insight sheds new light on the discussion. The problem of intentionality is not directly at issue so much as the problems of meaning and truthfulness are. But Searle’s approach could inspire one to think of settling down for a biologically viable account of naturalising semantic information. In doing so, we go beyond the scope of both of strategies mentioned above. Floridi attempted solving SGP in a non-derivative way and in connection to the AAs’ semantical capacities. In spite of Ross and Collier’s history of engagement with biological theories, Ladyman and Ross (2007) strategy was relying on ShannonWeaver’s and Barwise-Seligman syntactical or semisyntactical approaches. None of these proposals was centred on the biological constitution of the modellers. Under the circumstances, we may turn to the discipline of biosemiotics, instead of computer science, to deal with the problem. Incidentally, John Collier himself had a fine hand in broadening the scope of biosemiotics. Biosemiotics is an interdisciplinary study that integrates the fields of information theory, biology, psychology, and linguistics so as to deal with the issues of meaning and functionality in the world of living systems. Biosemiotics is a structural science8 to the extent that it makes use of mathematical and informational notions (such as codes and flow of information). But it also concerns the issue of the emergence of meaning, In speaking of Bthe structural sciences^ I follow the lead of Brunner, Taschacher, & Karsten, (2011).They classified the structural sciences (e.g. cybernetics and systems theory) as an independent field of the sciences which deal with the formal relations between the objects, irrespective of their content and nature. This structuralist aspect matters in our SR account. 8

M. D. Beni

functionality, and interpretation of the informational processing in the biological systems. In this part, I shall survey the possibility of a biosemiotics-based solution to the naturalisation problem. Here, the general insight is that a biosemiotics-based solution does not rely on any form of externalism, but it allegedly makes room for the agent’s (or even the molecular or the cellular unit’s) reliance on its own computational and perceptual resources. That being said, I have to add that there is a deep division among biosemioticians, between the code theorists and the Peircean proponents. The Periceans criticised the code model on account of its negligence of the importance of interpretants, especially in the field of endobiosemiotics which deals with issues of meaning and functionality at the molecular and cellular levels. In his BSigns without Minds,^ Collier (2014) claimed that the Peircean tradition could be developed to include the study of non-mental (including biological) signs. It could include the study of non-mental things that are signs for non-mental things. However, the non-mental signs, as Collier paraphrased, must play a genuine teleological or functional part for their bearer, in order to count as Peircean interpretants (Collier 2014, p. 184). Collier applied the idea of signs without minds to the field of biology when he argued that Bthere are non-cognitive things that do appear to have finality and the properties of representations. They use signs for their own purposes, but not through conscious activity^ (Collier 2014, p. 192). The implication is that there are genuine interpretants, but the interpretations are not the ones that a mind makes of a sign. Take the case of biological signalling. The signalling of an organism to oneself has been taken by Collier as a sign that does not serve the intentions of any conscious being, and hence it is totally mind-independent. The intentions of the signaller are not involved in the functional act, and the meaning that other organisms place on the signaller is not caused by its intentions. Therefore, there are external signs in biological systems that are not conscious, and Bevery non-mental biological sign is an entelechy, which is directly commensurate with its necessarily being functional^ (Collier 2014, pp. 193, 195).9 What adds to the biological processes their semiotic character is their Binterpretation in terms of survival,^ so that Bthe final interpretant would be the best function possible^ (Collier 2014, p. 195). Hence the possibility of genuine signs without minds. There are, however, reasons for thinking that Collier’s theory could not establish the mind-independence of the signs satisfactorily. 9

Perhaps it is worth mentioning that Griffiths and Stotz (2013 chapter 6) canvassed a similar view of genetic information. They assessed the possibility of assigning meaning to genetic codes in terms of adaptation for transmission of biological specificity from one generation to the next. The efficiency of this adaptation could be analysed in terms of an information channel using the mathematical theory of communication. The genetic transmission may be regarded as a way that the organisms use to send signals to their offspring. But, as the authors argued, it could not be claimed that it grounds the semantic content of those signals. At least it could be acknowledged that genetic signals do not have all the properties of signals in the human communication systems. The genetic codes should be understood in the sense of Crick information, which consists of the causal determination of the specificity of a biomolecule and comes with a mechanistic undertone. This adds up to the conclusion that using informational language in biology does not need to be associated with a semantic preformationism. Such an association may be based on confusion between evolutionary explanations and mechanistic ones. The authors concluded that the genetic code is a form of causal information, which is a causal relationship between one physical state and another one. The relation does not need to be more intentional than Grice’s Bnatural meaning.^ Thus, the coding problem could be solved on the basis of Shannon’s purely quantitative measure of information.

The Downward Path to Epistemic Informational Structural Realism

Collier’s theory has been offered as an improvement upon Charles S. Peirce’s theory, but the exegetical basis of Collier’s enterprise is not strong enough, and at times, it seems that he relied too much on possible implications of Peirce’s text. For example from Peirce’s remark BA Sign is a representamen of which some interpretant is a cognition of a mind,^ Collier concluded that there are triadic relations that are not signs because their interpretant is not a cognition (Collier 2014, p. 191). Collier seems to be aware of the limitations of the exegetical bases of his arguments. He handled the idea cautiously enough, and even hesitated to assert that the artefacts (e.g. weathervane) or even some natural objects (e.g. dust or leave blow, when used to determine the direction of the wind) could be signs without minds (Collier 2014, p. 192). Consequently, he confirmed that the mentioned examples are non-mental things that are signs only for the minds. Collier’s hesitation in approving the conclusion of a radical development of Peirce’s theory is understandable. There is, after all, a well-established branch of philosophy of biology that provides numerous arguments to underline the role of the interpreter in attributing meaning and functionality to biological systems. For example, Robert Cummins’ (Cummins 1975, 1985) teleological theory holds that teleological explanations are eligible only in accounting for the functioning of something that is a part of an artefact. That is to say, teleological explanations apply when there is some designer. The designer professes the belief that the relevant artefact will perform the function in question. In this scenario, the designer has devised the function in the system intentionally, in hopes that the system would perform its function in the relevant situation. It is possible to assume that the non-artefacts (i.e. biological systems) have functions too, but then again those functions could not be appealed to in a teleological story in the way of explaining the presence of the functions unless a mindful being ascribes the functions (Cummins, 1975). So, the cognitive agent which designs, interprets, and analyses complex systems ascribes functions to the parts of a complex (organic or inorganic) system. Even without being particularly sympathetic to Cummins’ systemic approach, one can observe that the role of intentional states of the spectator in ascribing functions in Normal conditions could not be easily overlooked. It is true that the existence of an alternative to Collier’s theory is not per se sufficient for arguing that signs could not be without minds. But it indicates that Collier’s construal overlooks the influence of purposes and explanatory goals of the analyst on the interpretation of the behaviour of interacting organisms. The analyst’s attempt at assigning meaning to the behaviour of the signaller could be a result of the analyst’s mentality, purposes, and explanatory goal. To establish the point that the biological non-mental signs do not owe their meaning to interpretations, Collier should prove that the intentions and goals of the interpreter do not affect her analysis of the biological system. He should have elaborated on the point that the signs and meanings of biological systems are not ascribed to them by a mindful being. But the requisite argument could not be found in Collier’s discussion. Under the circumstances, Collier’s view could lean towards ascriptionism. Ascriptionism bears some interesting resemblance to Dennett’s intentional stance which resisted to give way to realist reading in Sect. 3.2. So, despite its originality, the plausibility of Collier’s proposal is in need of further argumentation and evidence. But even if Collier had produced the requisite argument, the ramifications of his substantiated view would have been grave. His view could lead to monadic or other sophisticated forms of (Peircean) idealism. Collier’s theory indicates that non-mental

M. D. Beni

signs can play genuine teleological or functional roles for their bearers (which could be non-mental things). Therefore, the representational capacity (i.e. the capacity for using signs) can be attributed to either cognitive or non-cognitive beings. Therefore, different forms of monism and idealism may break in. If we assume that having representational capacity (for using signs) is the essential characteristic of having mental states (or conscious experience), Collier’s proposal can also lead to pan-psychism. This does not seem to be what Collier intended to say (given his emphasis on the non-mental nature of biological mechanisms). But the Peircean base of his thought easily allows for this construal. I do not suggest that these ramifications (monadic idealism, panpsychism) provide a decisive reason to rule out Collier’s theory. One may want to live with idealism and panpsychism. But they may ground a more suspicious attitude towards his theory. Let us attend to the code model alternative to Collier’s theory. Swimming against the Peircean tide, Barbieri denied that the interpreter and interpretations are indispensable parts of semiosis (Barbieri 2008, 2009, 2015). Thus, the Peircean triadic structure of semiosis (based on the distinction between object, representament, and interpretant) has been replaced with the Bcode-model^ which holds that a semiotic system consists of signs, meaning, and coding rules in addition to an indispensable codemaker. Functionality is used as an assumed trait of biological codes. According to this account, interpretive semiosis emerges on a rather recent stage of the history of life, along with the origins of animals. But even before generation of animals, the genetic codes had been actively contributing to manufacturing objects and organising them into functional systems. Meaning and signification were at work even at the molecular level. Signs and meanings can exist without a mind or an interpreter, but not without a codemaker or outside a code-making process (Barbieri 2009, p. 22). It is only at the interpretive level of the semiosis that the human mind is trusted with the job of code-making. At the more primary levels of code manufacturing semiosis (e.g. in protein synthesis and RNA’s splicing) and signalling semiosis (e.g. signal transduction), the codemaker is the organic system that is in charge of manufacturing processes and systematising them under an order which manifests the properties of functionality and meaningfulness. In protein synthesis (at the first stage of natural manufacturing semiosis), for example, a linear sequence of codons, i.e. nucleotides, could be encoded in different ways by a ribosome. The ribosome is the codemaker in this case, and the resulting amino acid is the reference or meaning of the codons. It is in this sense that Barbieri claimed the molecular process is merely codemaker-dependent, in spite of conveying both meaning and signification (Barbieri 2009, p. 22). Interpretive semiosis and mindfulness emerge at the more advanced stages of the life, but they are by-products of the same codemaking processes. Thus, the mind would be finally endowed with the ability of abduction, upon whose basis the animal learns to interpret the world by applying a Brule of the thumb^ that relates inputs and outputs to each other (ibid, 27). The gap between the semiosis in the world and interpretive and mindful activities could be traversed in this manner. At least in the early stages of formation of life, the process is not in need of an interpreting mind. Emergence of the mind and its upcoming interpretations are the results of the semiosis processes. Moreover, as I remarked earlier, Barbieri’s account is capable of finding clear information-theoretic expression. Coding process maps a domain of the entities (signs, a codon) onto the domain of other entities that could be understood as the meaning or referent of the denizens of the first domain.

The Downward Path to Epistemic Informational Structural Realism

It is a naturalistically plausible account of the relation of the informationally regimented structure of the theories to reality. This is the most persuasive strategy that we met thus far, but some reservations endure. In his account of biological functions, Collier argued that we have to rely on a robust sense of functionality in order to ground interpretants and interpretation. Otherwise, we would be left with purely mechanical sets of relations that could be reduced to causal relations alone (Collier 2011). In the face of such objection, the code-theorist could take one of the following options. One way is to forget about interpretants and their assumed function altogether, but then it could be claimed that code biology does not naturalise semantic information in the right way to connect representations to the world. The other option is to assume that the code theorist is actually smuggling interpretants in through the back door by invoking functionality without explaining what it is and why it is important. But this option does not go beyond an instrumentalist theory of meaning. The biosemiotics strategy, either in its Peircean or code-model guise, does not ground semantic information in a conclusively plausible manner. But as I have already remarked, the semantic information does not need to be conflated with the information that is at issue in ISR. Therefore, the negative conclusion does not need to indicate that the present shortcomings cannot be addressed appropriately, say, by working on a unifying strategy that combines the strong points of the code-model and the Peircean branches of the biosemiotics. Even so, I have to add that, in view of the reservations that have been spelt out in this paper, this last resort solution does not ground a fullblooded form of OSR. Although a unified view can ground informationally regimented structures in the mind-independent world, we do not need to go so far to assume that the informationally regimented structures capture all that there is. Despite the fact that this unified view still could not bear the semantical burden of a full-blooded (ontic) version of ISR, it can successfully underlie a modest version of ISR. In this paper, I offered a downward path to this form of ISR (see Sect. 3.3). That is to say, instead of offering independent positive arguments, I started from inflated metaphysics of OSR. But since its claims could not be defended naturalistically, we may adopt a more modest stance. The epistemic version of ISR, which is advocated in this paper, bears a resemblance to the argumentation that Worrall (1989) offered in his structuralist revival of Poincare’s (and Kant’s) views. He argued that ESR provides us with the best of both realist and empiricist worlds. It leans towards realism in assuming that our scientific theories provide us with knowledge of the world’s structure, albeit without going further to make any outright ontological commitments to structure of the mind-independent world.

5 Concluding Remarks In this paper, I argued that the inflated metaphysics of OSR could not be justified on the basis of the arguments of its advocates. We surveyed some attempts at naturalising information. But none of them could carry the burden of the metaphysical component of OSR. Under the circumstances, we may dispense with both naturalism and representationalism and try to defend OSR on the basis of a constructionist approach. But the main reservation is that constructionism throws out the baby with the bathwater and dissolves the problem of representation at the cost of realism. According to the

M. D. Beni

constructionist, scientific theories do not represent reality but construct knowledge through proactive interactions with an inexhaustible source of information. This is radically different from the familiar conception of scientific realism which is centred on the idea of approximately precise representation of reality. Alternatively, in this paper I outlined a downward path to EISR. EISR is an intermediary stance. It begins from outright informational OSR. But since the attempts at naturalising information could not carry the burden of the metaphysical component of OSR, EISR makes concession on a more modest form of informational SR. The modest form of ISR is in line with Worrall’s original formulation of SR. Compliance with Ethical Standards Conflict of Interest

The author declares that there is no conflict of interest.

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. Barbieri, M. (2008). Biosemiotics: a new understanding of life. Die Naturwissenschaften, 95(7), 577–599. https://doi.org/10.1007/s00114-008-0368-x. Barbieri, M. (2009). Three types of semiosis. Biosemiotics, 2(1), 19–30. https://doi.org/10.1007/s12304-0089038-9. Barbieri, M. (2015). Code biology. Cham: Springer International Publishing. https://doi.org/10.1007/978-3319-14535-8. Barwise, J., & Seligman, J. (1997). Information flow: the logic of distributed systems. Cambridge: Cambridge University Press. Beni, M. D. (2016). Epistemic informational structural realism. Minds and Machines, 26(4), 323–339. https://doi.org/10.1007/s11023-016-9403-4. Brunner, E. J., Taschacher, W., & Karsten, K. (2011). Selbstorganisation von Wissenschaft. Jena: VerlagIKSGaramond. Bueno, O. (2010). Structuralism and information. Metaphilosophy, 41(3), 365–379. https://doi.org/10.1111 /j.1467-9973.2010.01641.x. Collier, J. (2011). Explaining biological functionality: is control theory enough? South African Journal of Philosophy, 30(1), 53–62. https://doi.org/10.4314/sajpem.v30i1.64411. Collier, J. (2014). Signs without minds. In: V. Romanini & E. Fernández (Eds.) (pp. 183–197). Dordrecht: Springer Netherlands. doi: https://doi.org/10.1007/978-94-007-7732-3_10. Cummins, R. C. (1975). Functional analysis. The Journal of Philosophy, 72(20), 741. https://doi.org/10.2307 /2024640. Cummins, R. C. (1985). The nature of psychological explanation. Cambridge: MIT Press. Dennett, D. C. (1991). Real patterns. The Journal of Philosophy, 88(1), 27–51 Retrieved from http://links.jstor. org/sici?sici=0022-362X%28199101%2988%3A1%3C27%3ARP%3E2.0.CO%3B2-2. Dretske, F. I. (1981). Knowledge and the flow of information (first). Oxford: Blackwell Publishing Ltd.. 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. Esfeld, M., & Lam, V. (2008). Moderate structural realism about space-time. Synthese, 160(1), 27–46. https://doi.org/10.1007/s11229-006-9076-2. 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. Floridi, L. (2005). Is semantic information meaningful data. Philosophy and Phenomenological Research, LXX(2), 351–370. https://doi.org/10.1111/j.1933-1592.2005.tb00531.x. Floridi, L. (2008). A defence of informational structural realism. Synthese, 161, 219–253. https://doi. org/10.1007/s11229-007-9163-z.

The Downward Path to Epistemic Informational Structural Realism Floridi, L. (2009). Against digital ontology. Synthese, 168(1), 151–178. https://doi.org/10.1007/s11229-0089334-6. Floridi, L. (2011). The philosophy of information. Oxford: Oxford University Press. https://doi.org/10.1093 /acprof:oso/9780199232383.001.0001. Floridi, L. (2017). A plea for non-naturalism as constructionism. Minds and Machines, 27, 1–17. https://doi. org/10.1007/s11023-017-9422-9. van Fraassen, B. C. (1980). The scientific image. Oxford: Oxford University Press. https://doi.org/10.1093 /0198244274.001.0001. French, S. (2014). The structure of the world metaphysics and representation. Journal of Chemical Information and Modeling (Vol. 53). doi: https://doi.org/10.1017/CBO9781107415324.004. 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. Griffiths, P., & Stotz, K. (2013). Genetics and philosophy: an introduction. Cambridge. Retrieved from http://www.cambridge.org/ir/academic/subjects/philosophy/philosophy-science/genetics-and-philosophyintroduction?format=PB&isbn=9780521173902#c74UZwIbiF9H7E0M.97. Harnad, S. 1990. “The Symbol Grounding Problem.” Physica D: Nonlinear Phenomena 42(1–3): 335–46. https://doi.org/10.1016/0167-2789(90)90087-6. Ladyman, J., Douven, I., Horsten, L., & Fraassen, B. (1997). A defence of van Fraassen’s critique of abductive inference: reply to Psillos. The Philosophical Quarterly, 47(188), 305–321. https://doi.org/10.1111/14679213.00061. Ladyman, J., & Ross, D. (2007). Every Thing Must Go. Oxford: Oxford University Press. https://doi. org/10.1093/acprof:oso/9780199276196.001.0001. Long, B. R. (2014). Information is intrinsically semantic but alethically neutral. Synthese, 191(14), 3447– 3467. https://doi.org/10.1007/s11229-014-0457-7. Morganti, M. (2011). Is there a compelling argument for ontic structural realism? Philosophy of Science, 78(5), 1165–1176. Muller, F. A. (2011). Withering away, weakly. Synthese, 180(2), 223–233. https://doi.org/10.1007/s11229009-9609-6. Psillos, S. (2000). The present state of the scientific realism debate. The British Journal for the Philosophy of Science, 51(4), 705–728. https://doi.org/10.1093/bjps/51.4.705. Psillos, S. (2001). Is structural realism possible? Philosophy of Science, 68(S3), S13–S24. https://doi. org/10.1086/392894. Searle, J. R. (1980). “Minds, Brains, and Programs.” Behavioral and Brain Sciences 3(3). Cambridge University Press: 417. https://doi.org/10.1017/S0140525X00005756. Shannon, C. E., & Weaver, W. (1949). The mathematical theory of communication. Urbana: University of Illinois Press. Taddeo, M., & Floridi, L. (2005). “Solving the Symbol Grounding Problem: A Critical Review of Fifteen Years of Research.” Journal of Experimental and Theoretical Artificial Intelligence 17(4): 419–45. 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. Whittle, A. (2006). On an argument for humility. Philosophical Studies, 130(3), 461–497. https://doi. org/10.1007/s11098-004-5751-7. 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.