A Cognitive Computation Fallacy? Cognition, Computations and Panpsychism
Summary (3 min read)
Introduction
- In operation, such computationally defined neurons effectively sum up their input and compute a complex nonlinear function on this value; output information being encoded in the mean firing rate of neurons, which in turn exhibit narrow functional specialisation.
- In the literature there exist numerous examples of such learning rules and architectures, more or less inspired by varying degrees of biological plausibility; early models include [3–6].
- From this followed the functional specialization paradigm, mapping different areas of the brain to specific cognitive functions.
- 9] and (iii) an implicit non-reductive functionalism with respect to consciousness [10].
The CTM
- The CTM occupies one part of the spectrum of representational theories of mind (RTM).
- Harnish [7] observes that the RTM entails: – Cognitive states are relations to mental representations which have content.
- – Psychological states are multiply realisable in different organisms.
- Putnam’s 1967 conclusion is that the best explanation of the joint multiple realisability of TMs and psychological states3 is that TMs specify the relevant functional states and so specify the psychological states of the organism; hence by this observation Putnam makes the move from ‘the intelligence of computation to the computational theory of intelligence’ [7].
- – A cognitive state is a state [of mind] denoting knowledge; understanding; beliefs, etc. – Cognitive processes—changes in cognitive states—are computational operations on these computational representations.
The Problem of Consciousness
- The term ‘consciousness’ can imply many things to many different people.
- Its negative conclusion is that computation is neither necessary nor sufficient for cognition; its positive conclusion suggests that the adoption of a new metaphor may be helpful in addressing hard conceptual questions related to consciousness and understanding.
- An analogy is with the application of Newtonian physics and Quantum physics—both useful descriptions of the world, but descriptions that are most appropriate in addressing different types of questions.
- The idea that the appropriately programmed computer really is a mind, and was eloquently suggested by Chalmers (ibid).
- This asserts that, ‘‘given any system that has conscious experiences, then any system that has the same fine-grained functional organization will have qualitatively identical experiences’’.
The Problem of Computation
- It is a commonly held view that ‘there is a crucial barrier between computer models of minds and real minds: the barrier of ness’ and thus that ‘information-processing’ and ‘phenomenal experiences’ are conceptually distinct [23].
- In the following sections I briefly review two wellknown arguments targeting computational accounts of cognition from Penrose and Searle, which together suggest computations are neither necessary nor sufficient for mind.
- Arguments based on Gödel’s first incompleteness theorem—initially from Lucas [25, 26] were first criticised by Benacerraf [27] and subsequently extended, developed and widely popularised by Penrose [28–31]—typically endeavour to show that for any such formal system F, humans can find the Gödel sentence Gð gÞ whilst the computation/machine (being itself bound by F) cannot.
- Now in response to the questions in Chinese and English there are two subsystems—the native English speaking Searle and the internalised Searle-in-the-Chinese-room—but all the same he [Searle] continues to understand nothing of Chinese, and a fortiori neither does the system, because there is nothing in the system that is not just a part of him.
Discrete State Machines
- Turing defined DSMs as ‘‘machines that move in sudden jumps or clicks from one quite definite state to another’’ and explained that modern digital computers fall within the class of them.
- Turing demonstrated that such a device, which continually jumps through a linear series of state transitions like clockwork may be implemented by a simple discrete-position-wheel that revolves through 120 intervals at each clock tick.
- Note, after Chalmers, that the discrete position wheel machine described above will only implement a particular execution trace of the FSA and Chalmers remains unfazed at this result because he states that input-less machines are simply an ‘‘inappropriate formalism’’ for a computationalist theory of mind [32].
Objections
- As the experimental setup is precisely the same for experiment (2) as for experiment (1) the computationalist must continue to claim that the robot continues to instantiate appropriate phenomenological states over this period and it is clear that a posteriori knowledge of the system input does not impact this claim.
- It is apparent that under mapping A (see Table 4), the gate X computes the logical AND function.
- The Objection From Randomness Superficially the DwP reductio only targets DSMs; it has nothing to say about the conscious state of suitably engineered Stochastic Automata [60].
- At the ‘Toward a Science of Consciousness’ conference in Tucson 2006 Ron argued that as the authors morph between R1 and R2 with the deletion of each conditional non-entered state sequence substantive physical differences between R1 and R2 will emerge.
Is Counterfactual Sensitivity Essential to a Computational Account of Cognition?
- The supervenience thesis tells us that, if the authors introduce into the vicinity of the system an entirely inert object that has absolutely no causal or physical interaction with the system, then the same activity will support the same mode of consciousness.
- So despite Bishop’s claim, if R1 and R2 differ in their counterfactual formal properties, they must differ in their physical properties.
- In each of the following experiments the robots are instructed to report the colour of a large red square fixed in the centre of their visual field.
- Next the virtual robot software is re-complied using two slightly different partial evaluation compilers [65].
- However the reductio targets computationalism—the formal abstraction and instantiation of consciousness through appropriate DSMs (and/or their stochastic variants); the DwP reductio does not target continuous [dynamic] systems or identity theories (where conscious properties of the system are defined to be irreducible from the underlying physical agent–environment system).
Are These A Priori Critiques of the Computational Metaphor too Strong?
- Interestingly, as this form of compile time partial evaluation process cannot be undertaken for the real robot, the DwP reductio strictly no longer holds against it; however, this does not help the computationlist as any putative phenomenal states of the real robot have now become tightly bound to properties of the real-world agent/environment interactions and not the mere computations.
- There are two responses to this question, one weak and one strong.
- The first—the weak response—emerges from the Chinese room and DwP reductio.
- Hence Searle’s famous observation that ‘‘… the idea that computer simulations could be the real thing ought to have seemed suspicious in the first place because the computer is not confined to simulating mental operations, by any means.
- Both of the above responses accommodate results from computational neuroscience, but clearly both also highlight fundamental limitations to the computational metaphor.
So what Is Cognition, If Not Computation?
- In contrast to computation, communication is not merely an observer-relative anthropomorphic projection on reality, as even simple organisms (e.g. bacteria) communicate with each other or interact with their environment.
- Thus the new metaphor— cognition as communication—is sympathetic to modern post-symbolic, anti-representationalist, embodied, enactive accounts of cognition such as those from Brooks [70], Varela [19], O’Regan [15], Thompson [71] and Bishop and Nasuto [13].
Conclusion
- All matter, from the simplest particles to the most complex living organisms, undergo physical processes which in most sciences are not given any special interpretation.
- In neuroscience, and in connectionism, it is assumed that neurons and their systems possess special computational capabilities; this is equivalent to claiming that a spring, when extended by a moderate force, computes its deformation according to Hooke’s law.
- But at heart he follows an extremely simple line of reasoning: consider an idealised analogue computer that can add two reals (a, b) and output one if they are the same, zero otherwise.
- I would like to thank the reviewers for the many helpful comments I received during the preparation of this article.
Did you find this useful? Give us your feedback
Citations
15 citations
Cites background from "A Cognitive Computation Fallacy? Co..."
...problematic, and even detrimental to computational analyses of cognitive systems (see, for example, Bishop [2009]; Sprevak [2010]; Shagrir [2012]). In the debate on the metaphysical nature of computation, whether or not a given account of computation is able to ‘resolve’ the indeterminacy of computation problem has been considered a litmus test for its adequacy. The notorious case study from Boolean engineering for analysing this phenomenon has been the duality of two simple Boolean gates: a two-input, single-output AND-gate and a two-input, single-output OR-gate. There is nothing special about these two gates: the phenomenon occurs in twelve out of sixteen two-input, single-output Boolean gates (XOR and XNOR, NAND and NOR, and so on). Two noteworthy, recent discussions of this phenomenon are due to Dewhurst ([2018]) and Lee ([forthcoming]), both of which centre on the question of computational individuation....
[...]
...199; Shagrir [2001], [2012]; Buechner [2008]; Bishop [2009]; Sprevak [2010]; Fresco [2015]; Piccinini [2015]; Miłkowski [2017]; Coelho Mollo [2018])....
[...]
...problematic, and even detrimental to computational analyses of cognitive systems (see, for example, Bishop [2009]; Sprevak [2010]; Shagrir [2012])....
[...]
14 citations
Cites background from "A Cognitive Computation Fallacy? Co..."
...In [7] Bishop reviews three arguments (summarised herein) that purport to show that computations are not sufficient for cognition; for example, that the execution of a computational connectionist simulation of the brain cannot instantiate genuine understanding or phenomenal consciousness (qua computation) and hence that there are limits to the use of the computational explanations in cognitive science....
[...]
...Thus a corollary of O’Regan’s move to close the absolute gap [by a refinement of merely abstract formal processes] implies that it - in common with the other two “VR robot systems” highlighted above - would be phenomenally conscious purely in virtue of its execution of an appropriate computer program; and hence that it would be vulnerable to the various critiques of computationalism [7], in so far as these critiques hold at all....
[...]
13 citations
13 citations
Cites background from "A Cognitive Computation Fallacy? Co..."
...Some, for example Shagrir (2020), Bishop (2009) and Sprevak (2010), appeal to semantic features of the system to render it determinate what computation is being performed....
[...]
...…computation (Fresco et al. 2016) have been discussed by diverse authors—pre-eminently Shagrir (2001, 2020)—and also (in chronological order) Dennett (1978, 2013), Block (1990), Sorensen (1999), Bishop (2009), Sprevak (2010), Fresco (2015), Piccinini (2015), Coelho Mollo (2017), and Dewhurst (2018)....
[...]
11 citations
References
19,800 citations
Additional excerpts
...biological plausibility; early models include [3–6]....
[...]
12,923 citations
"A Cognitive Computation Fallacy? Co..." refers background in this paper
...After Hubel and Wiesel [2] this view of the neuron as a specialised feature detector has become treated as established doctrine....
[...]
8,434 citations
Additional excerpts
...biological plausibility; early models include [3–6]....
[...]
8,134 citations