Epistemology of Analogue Simulation

Epistemology of Analogue Simulation
University of Geneva
Tuesday 10 September
9:20am to 5:40pm

We are pleased to announce registrations are now open for the workshop "Epistemology of Analogue Simulation" to be held at the University of Geneva on 10 September, 2019. The workshop is jointly organised by the University of Queensland and the University of Geneva.

This workshop is open for all to attend. Registration is free, but required for catering purposes.

The workshop venue is room M1150, Uni Mail, Boulevard du Pont-d'Arve 40, 1205 Geneva.

Schedule:

9:20 - Welcome

9:30 - On the Universality of Hawking Radiation and its role in analogue experiments - Patricia Palacios (Salzburg) & Sean Gryb (Groningen)

10:30 - What is So Special About Analogue Simulations? - Francesco Nappo (UNC, Chapel Hill)

11:10 - Coffee

11:40 - Empirical Evidence or an Amusing Feat of Engineering? The problem with analogue simulation - Grace Field (Cambridge)

12:20 - On the Limits of Experimental Knowledge - Peter Evans (Queensland)

13:20 - Lunch

14:30 - Digital is Analogue: analogue experiments, in silico experiments and simulations - Michal Hladky (Geneva)

15:10 - Computer and analog simulation: Are they really any different? - Florian Boge (Aachen)

16:10 - Coffee

16:40 - What can we learn from *quantum* analogue simulations? - Dominik Hangleiter (FU Berlin)

17:40 - End

Analogue simulation techniques have taken on increasing significance in experimental science in recent years. In particular, analogue black hole experiments appear to provide insight into black hole phenomena, such as Hawking radiation, and analogue quantum simulation appears to provide a powerful computational tool for analysing intractable quantum systems. This workshop aims to explore the epistemology of analogue simulation, considering (but not limited to) the following sorts of questions: What epistemic license does analogue simulation warrant? In what ways, if at all, does the epistemology of analogue simulation differ from conventional experiment or computer simulation? In what ways, if at all, is the epistemology of analogue quantum simulation idiosyncratic compared to other forms of analogue simulation?

All questions concerning the workshop should be directed to the workshop convenors:
Peter Evans (University of Queensland): [email protected]
Karen Crowther (University of Geneva): [email protected]

Abstracts:

Patricia Palacios (Salzburg) & Sean Gryb (Groningen)

On the Universality of Hawking Radiation and its role in analogue experiments

A physically consistent semi-classical treatment of black holes requires universality arguments to deal with the `trans-Planckian' problem where quantum spacetime effects appear to be amplified such that they undermine the entire semi-classical modelling framework. We evaluate three families of such arguments in comparison with Wilsonian renormalization group universality arguments found in the context of condensed matter physics. Our analysis is framed by the crucial distinction between robustness and universality. Particular emphasis is placed on the quality whereby the various arguments are underpinned by `integrated' notions of robustness and universality. Whereas the principal strength of Wilsonian universality arguments can be understood in terms of the presence of such integration, the principal weakness of all three universality arguments for Hawking radiation is its absence. We finally analyze the consequences of these results for the validity of analogue black holes.

Francesco Nappo (UNC, Chapel Hill)

What is So Special About Analogue Simulations?

Dardashti, Hartmann, Thebault and Winsberg (2018; henceforth: DHTW) and Crowther, Linneman and Wütrich (2019; henceforth: CLW) have taken opposite sides on the issue whether experiments concerning 'dumb holes' in fluids and Bose-Einstein condensates can provide incremental confirmation of yet untested hypotheses about properties of black holes in quantum gravity: briefly put DHTW answer 'yes' and CLW answer 'no'. DHTW's reason for answering 'yes' is that the experiments provide some confirmation to the hypothesis that black holes exhibit the 'same kind' of physical system as fluids under very special conditions, which in turn is supposed to constitute evidence that black holes really are such 'radiating' objects as predicted by Hawking and others. CLW deny this: in their view, these analogue experiments have no business telling us anything informative about black holes, since the latter are inaccessible to us. What I am going to suggest in this paper is that they are both wrong: CLW are, I think, too pessimistic about what we can learn from analogue experiments, whereas DHTW give the right answer but are just wrong about what the experiments tell us.

Grace Field (Cambridge)

Empirical Evidence or an Amusing Feat of Engineering? The problem with analogue simulation

Analogue simulation claims to use the behaviour of an experimentally accessible 'source' system to learn about the behaviour of an experimentally inaccessible 'target' system. But the status of that inferential link between the source and target systems has - not surprisingly - proved controversial. Dardashti et al. (2017; henceforth, 'DTW') and Dardashti et al. (2019; henceforth, 'DHTW') provide a philosophical defence of analogue simulation as "a novel form of scientific inference with the potential to be confirmatory" (DTW, 55). Crowther et al. (2019; henceforth, 'CLW') counter that analogue simulation, at least as presented by DTW and DHTW, suffers from vicious circularity. These three works therefore establish a debate that requires further adjudication, which will be the aim of this paper.

To begin, I present a summarized, synthesized, and clarified account of analogue simulation (section II). Based on that account, I go on to draw my own conclusions on the strengths and limitations of analogue experimentation. In section III.i, I explain why the most basic version of analogue simulation suffers from circularity. In section III.ii, I show that DTW does suffer from the circularity problem outlined in III.i: CLW's criticism of DTW is convincing. However, I point out that the direct circularity in DTW is not present in DHTW; therefore, it is not fair of CLW to imply that their criticisms of DTW can apply equally well as criticisms of DHTW. DHTW make a suggestion that is not directly circular, and at first sight seems as if it might be able to rescue the confirmatory power of analogue simulation.

Ultimately, however, I will conclude in section III.iii that analogue simulation fails to provide confirmation for hypothesized features of the target system. It fails for a reason identified by CLW. But where CLW state that reason by referring back to their criticisms of DTW, I provide a separate argument (section IV). Namely, DHTW's attempt to rescue analogue simulation requires us to be able to identify conditions that support the adequacy of both the source and target models as descriptions of their respective systems. Their attempt fails because the circumstances in which we need analogue simulation are precisely the circumstances in which we have no ability to identify those kinds of conditions.

Peter Evans (Queensland)

On the Limits of Experimental Knowledge

This talk utilises the contrast between conventional experiments and analogue experiments as a diagnostic tool to isolate the key factors that allow conventional experiments to be awarded a unique epistemic status within science. The focus of the talk is two specific case studies probing the relationship between theory and evidence: the first probes the evidence for the model of stellar nucleosynthesis (an exemplar of evidence from conventional experimentation); and the second probes the evidence for black hole models of Hawking radiation (an exemplar of evidence from analogue experimentation). This talk identifies the 'inductive triangulation' of evidence as a strategy employed in observational astrophysics in support of theories regarding unobservable, unmanipulable, and inaccessible target phenomena. We argue that the probative value of analogue experiments in part depends upon scientists ability to combine them with other lines of evidence, and so enable the strategy of inductive triangulation in support of black hole models of Hawking radiation.

Michal Hladky (Geneva)

Digital is Analogue: analogue experiments, in silico experiments and simulations

Several debates in philosophy of science focus on the epistemic status of novel scientific practices. The epistemic power of analogue experiments (Dardashti et al. 2019; Crowther, Linnemann, and Wüthrich 2019), in silico experiments, computer simulations is compared with the gold standard of empirical sciences - experiments (Guala 2002; Morgan 2003; Parker 2009; Winsberg 2009; Roush 2017). Before evaluating any epistemic question, it is necessary to specify what one means by these notions.

First, contrary to what is often presupposed (Guala 2002), formal analyses can show that these terms do not denote nonoverlapping methods. For instance, a form of experimental manipulation often is an integral part of simulations (for opposing view see Beisbart 2017). Furthermore, it is not clear how to understand the expressions 'analogue simulation', 'analogue experiment' and even 'analogue computation'. In case of computer simulations, it is crucial to specify what is meant by 'computer'. Usually, one refers to digital electronic computers based on von Neumann architecture. But quantum and neuromorphic architectures cannot be fully described by the classical theory of computation. More faithful accounts are based on physical descriptions of such systems and their mapping to targets. Colloquially, the distinction is made in terms of digital and analogue processing. But as the old engineering proverb states, digital is analogue. It is preferable to characterise these two computational alternatives in terms of discrete and continuous computational states (Maley 2011, 2018; Stukelj 2019) and reserve the predicate 'analogue' to a form of reasoning.

Second, with a reconstruction of inferences typical for experimental and simulation contexts, one can demonstrate that those related to experiments introduce less premisses and are therefore more powerful. This holds however only as a ceteris paribus claim and in particular instances, arguments based on simulations can yield conclusions that have stronger justification than those based on experiments, similarly to Parke's (2014) conclusions.

Finally, it is possible to adopt a form of perspectivalism and describe experiments as simulations and vice-versa. Such an approach can explain the conflicting intuitions about in silico experiments, analogue experiments and simulations. But as not all perspectives are equal, this comes at the cost of introducing less plausible premisses and can undermine the conclusions about target phenomena.

Florian Boge (Aachen)

Computer and analog simulation: Are they really any different?

In their recent paper on confirmation via analog simulation (BJPS, 2017, 68(1): 55–89), Dardashti, Thébault and Winsberg casually mention that, on their account, both analog and computer simulation are just instances of simulation. The only real difference, according to them, rests in the fact that in computer simulation, the simulating system is a programmed digital computer, whereas in analog simulation it is not. However, the arguments offered for this claim in the paper remain rather underdeveloped, as the focus of the paper is a different one. In my talk, I will dig a little deeper, and offer an account of how and in what sense analogies figure in the case of computer simulation, just as much as they do in analog simulation. For that purpose, I first introduce an account of analogy as structure matching via approximate homomorphic mappings, of which the isomorphisms featuring in analog simulation are just a special case. I then use a toy example to demonstrate how we can in principle identify such a mapping between a programmed digital computer and the target of a simulation. Subsequently, I consider the impact of a rival account recently advocated by Beisbart (EJPS, 2018, 8(2): 171–204), and provide some reasons for doubt.

Dominik Hangleiter (FU Berlin)

What can we learn from *quantum* analogue simulations?

Whereas classical analogue simulators have largely been superseded by classical computer simulation, the surging technological advances in coherently manipulating individual quantum degrees of freedom make the idea of using such bespoke systems to simulate other quantum systems one of the most promising applications of near-term quantum devices. This idea is captured by the term "analogue quantum simulation", which is used to refer to a wide range of epistemic activities. But what makes an analogue simulation specifically *quantum*? What exactly is the epistemic goal of analogue quantum simulation: is it a novel mode of inference or does it reduce to traditional ones such as analogical argument, computer simulation, or experimentation? And finally: What are the conditions for success or failure of analogue quantum simulation? In this talk, I will approach those questions and sketch a prospectus to the epistemology of analogue quantum simulation by analyzing a few informative case studies. Based on those I will first discuss what is really "quantum" about quantum simulation. In a second step, I will then argue that it is helpful to introduce a distinction between two types of analogue simulation -- computation and emulation -- to differentiate between the types of understanding obtained in different contexts. Finally, I will provide concrete failure conditions and develop respective notions of certification as epistemic norms for successful analogue quantum emulation and computation.