For as long as astrophysicists have considered the large-scale structure of the cosmos, discoveries on the subject have been taken to provide critical empirical insights relevant to theorizing in fundamental physics. This close connection between the two subjects is most familiar in the context of the relativistic hot 'Big Bang' model of the expanding universe that first rose to prominence in the 1930s, and which later developed into the standard Lambda-CDM model familiar today. In that hot 'Big Bang' model, the developmental origins of large-scale cosmic structure present today in our observable universe are understood in terms of high-energy fundamental physical processes far beyond our reach. And in Lambda-CDM, this role for fundamental physics beyond our reach is supplemented by additional roles, via theorizing about a dark sector to be incorporated into future physics. But through history, competitor accounts of the large-scale structure of the cosmos have likewise included empirical upshots that were to be taken as clues about new high-energy fundamental physics: witness the particle basis for the 'C-field' in Hoyle’s approach to Steady State cosmology in the 1950s, and even the earlier call in the 1930s to revise our fundamental understanding of classical gravitational dynamics within the context of Milne’s 'kinematic' model of the receding nebulae. In light of this consistent close connection between the empirical claims of large-scale cosmology and ongoing theorizing in fundamental physics, it is unsurprising that researchers interested in quantum gravity have often tried to mine the Lambda-CDM model for empirical clues that pertain specifically to their own domain of theorizing: looking for explicit cosmic imprints of quantum gravity. In this talk, after arguing in brief for the historical point above, I will then carefully pull apart (and critically assess) what I take to be two distinct recent programs of empirical QG research that fit the historical pattern. The first program treats the dynamics of Lambda-CDM as an autonomous structural theory of our present-day observable universe at the largest accessible scales (in terms of a nearby portion of the large-scale cosmos), to explore how detailed processes in quantum gravity might lead to 'trans-Planckian' physics in that same target system, which almost mimics familiar expectations about the large-scale cosmos derived from known fundamental physics. The second program treats Lambda-CDM, the model, as a low-energy effective description of a future model of quantum cosmology, so as to consider whether a sustained commitment to the descriptive accuracy of the former may ultimately be constraining on the topic of how to quantize gravity (i.e. in order to eventually construct the latter). I will conclude by noting a curious feature of the new 'trans-Planckian censorship' conjecture that has been advertised as a general principle governing possible low-energy physical descriptions in a universe where gravity is quantized: that the two empirical programs just pulled apart would seem, perhaps, to collapse into one --- albeit contingent on very particular speculations about what may come of our future understanding of the quantum nature of gravity within our present-day observable universe.

Although quantum gravity is often described as empirically inaccessible, in fact astrophysics and cosmology teem with situations in which both gravitational and quantum-mechanical effects are relevant, and so we have abundant observational constraints on quantum gravity at energy levels low compared to the Planck scale. That evidence supports (to a variable degree) the description of quantum gravity as an effective field theory version of general relativity, breaking down at Planckian energies. I briefly present this theory, review its empirical support and its problems (including the cosmological constant problem) and consider the prospects for gaining empirical evidence for quantum gravity beyond the low-energy regime.

It has long been thought that observing the effects of quantum gravity is effectively impossible, since gravity is so much weaker than other forces: consider, for instance the utterly insensible gravitational attraction of a magnet, compared with the very sensible magnetic force it exerts. But by drawing on ideas from 'quantum information theory' (QIT), and on recent experimental advances in quantum mechanics and in observing tiny gravitational fields, Bose et al (2017), and Marletto and Vedral (2017) have shown how, in principle, weak gravitational fields might have detectable quantum effects. This work has attracted a great deal of interest, as such 'BMV experiments' are tantalizingly close to current experimental physics; many experimentalists are working on the real possibility that the predicted effect could be measured in the next few years, even though the experiment would be one of the most delicate ever undertaken. But its significance would be a momentous advance in the study of quantum gravity. There are a number of conceptual issues to unpack regarding this proposal: First, in what sense would the BMV experiment count as an observation of the 'quantum nature' of gravity? First, there are theoretical issues: different, seemingly equally reasonable, theoretical commitments can lead to rather different understandings of the meaning and significance of the experiment. On the one hand, the BMV argument assumes that gravity should be modeled as a dynamical system. On the other, a gauge theoretic approach to quantizing gravity explains the effects in terms of a non-dynamical gauge constraint. Then, a further issue is the comparison with previous experiments which combined quantum and gravitational effects; what new information would we obtain from the BMV experiment, and in what senses would we gain greater practical control over quantum gravity? The philosophical issues thus concern the interpretation of physical theory, and the nature and role of experiment in science: both important topics within philosophy of physics and philosophy of science. Finally, moreover, QIT is in the first place a specific formulation of quantum mechanics, but it can involve further more specific assumptions. In the analysis of the BMV experiment in particular, one has to stipulate what it is for a system to be 'classical' rather than 'quantum'. Clearly this is an important, contentious, but undertheorized question in the philosophy of physics literature. This talk will outline the idea of the BMV experiment, and address the philosophical issues that it raises: (1) arguing that ultimately the *the* meaning of the 'quantum nature’ of a system is arbitrary to a certain extent; (2) explaining how the BMV experiment would both give an observation of the quantum nature of gravity in a deeper sense, and require more robust control over it, than previous experiments; (3) clarify the stakes in the assumptions of the QIT theorem.

Analogue experiments have attracted interest for their potential to shed light on inaccessible domains. In 1981, Unruh found a striking mathematical analogy between the propagation of light waves near a black hole and the propagation of sound in fluids. In fact, a number of distinct such 'analogue' systems can be found, from hydrodynamical systems to Bose-Einstein condensates. The remarkable discovery of an analogy between black holes and 'dumb holes' in fluids ('swallowing' sound) or Bose-Einstein condensates has spawned a rich literature exploring the emerging field of 'analogue gravity'. Moreover, it has led to an active experimental field of studying such analogue systems in labs, culminating in the observation of analogue Hawking radiation in Bose-Einstein condensates. Analogue gravity thus naturally leads to the question of whether such analogue models can confirm the existence of (gravitational) Hawking radiation in astrophysical black holes. Can we learn anything about black holes from these analogue models? More generally, can analogue models confirm hypotheses regarding inaccessible target systems? While Dardarshti et al. have argued that analogue gravity can indeed confirm gravitational Hawking radiation, Crowther et al. have criticized their argument as circular: in order to ascertain that the analogue model and black holes in fact fall into the same universality class, one must assume that black holes are adequately described by the modelling framework from which Hawking radiation is derived---but this is precisely what was to be confirmed. Analogue experiments whose target systems are inaccessible in some epistemically relevant sense generally suffer from this weakness: they must assume the physical adequacy of the modelling framework in order to underwrite the analogy to the accessible lab system when it is often precisely this adequacy which is at stake. Recently, Evans and Thébault (and to some degree Field) have equated concerns about this circularity with general inductive scepticism. Extrapolating from lab experiments on fluids to inaccessible black holes, is, according to them, no different in principle from inferring from today's experimental results to those of tomorrow. In support of this claim, they enlist the example of stellar nucleosynthesis, i.e., of reactions transforming hydrogen to helium in the interior of main sequence stars such as our sun. These reactions are unmanipulable and (photonically) inaccessible to us; nevertheless, astrophysical observations and terrestrial experiments in nuclear physics largely confirm the theory of stellar nucleosynthesis. This argument raises the subtle and rich question of under what circumstances we can think of two systems as being of the same or of a different 'type', i.e., under what conditions can experiments on one system be considered confirmatory of a theory on another system. This talk will show that this question is relevantly different from Hume's general inductive scepticism, particularly if it concerns inferences to inaccessible target systems. When we make inferences from experimental results of analogue systems to inaccessible target systems, we require the substantial assumption that our experimental and target systems are of the same type. It is for this assumption that we need independent support.