We discuss scientific realism from the perspective of modern cosmology, especially primordial cosmology: i.e. the cosmological investigation of the very early universe. We first state our allegiance to scientific realism, and discuss what insights about it cosmology might yield, as against "just" supplying scientific claims that philosophers can then evaluate. In particular, we discuss: the idea of laws of cosmology, and limitations on ascertaining the global structure of spacetime. Then we review some of what is now known about the early (...) universe : meaning, roughly, from a thousandth of a second after the Big Bang onwards. The rest of the paper takes up two issues about primordial cosmology, i.e. the very early universe, where "very early" means, roughly, much earlier than one second after the Big Bang: say, less than 10^{-11} seconds. Both issues illustrate that familiar philosophical threat to scientific realism, the under-determination of theory by data---on a cosmic scale. The first issue concerns the difficulty of observationally probing the very early universe. More specifically, the difficulty is to ascertain details of the putative inflationary epoch. The second issue concerns difficulties about confirming a cosmological theory that postulates a multiverse, i.e. a set of domains each of whose inhabitants cannot directly observe, or otherwise causally interact with, other domains. This again concerns inflation, since many inflationary models postulate a multiverse. For all these issues, it will be clear that much remains unsettled, as regards both physics and philosophy. But we will maintain that these remaining controversies do not threaten scientific realism. (shrink)

It is commonly argued that an undesirable feature of a theoretical or phenomenological model is that salient observables are sensitive to values of parameters in the model. But in what sense is it undesirable to have such ‘fine-tuning’ of observables? In this paper, we argue that the fine-tuning can be interpreted as a shortcoming of the explanatory capacity of the model: in particular it signals a lack of a particular type of explanatory depth. The aspect of depth that we probe (...) relates most closely to a lack of sensitivity to changes in parameters associated with such models. In support of this argument, we develop a schema—for models that arise broadly in physical settings—that quantitatively relates fine-tuning of observables to a lack of depth of explanations based on these models. We apply our schema in two different settings in which, within each setting, we compare the depth of two competing explanations. The first setting involves explanations for the Euclidean nature of spatial slices of the universe today: in particular, we compare an explanation provided by the big-bang model of the early 1970s with an explanation provided by a general model of cosmic inflation. The second setting has a more phenomenological character, where the goal is to infer from a limited sequence of data points, using maximum entropy techniques, the underlying probability distribution from which these data are drawn. In both of these settings we find that our analysis favors the model that intuitively provides the deeper explanation of the observable of interest. We thus provide an account that relates two ‘theoretical virtues’ of models used broadly in physical settings—namely, a lack of fine-tuning and explanatory depth—and argue that finely tuned models sacrifice explanatory depth. (shrink)

In the absence of a fundamental theory that precisely predicts values for observable parameters, anthropic reasoning attempts to constrain probability distributions over those parameters in order to facilitate the extraction of testable predictions. The utility of this approach has been vigorously debated of late, particularly in light of theories that claim we live in a multiverse, where parameters may take differing values in regions lying outside our observable horizon. Within this cosmological framework, we investigate the efficacy of top-down anthropic reasoning (...) based on the weak anthropic principle. We argue contrary to recent claims that it is not clear one can either dispense with notions of typicality altogether or presume typicality, in comparing resulting probability distributions with observations. We show in a concrete, top-down setting related to dark matter, that assumptions about typicality can dramatically affect predictions, thereby providing a guide to how errors in reasoning regarding typicality translate to errors in the assessment of predictive power. We conjecture that this dependence on typicality is an integral feature of anthropic reasoning in broader cosmological contexts, and argue in favour of the explicit inclusion of measures of typicality in schemes invoking anthropic reasoning, with a view to extracting predictions from multiverse scenarios. (shrink)

In extracting predictions from theories that describe a multiverse, we face the difficulty that we must assess probability distributions over possible observations, prescribed not just by an underlying theory, but by a theory together with a conditionalization scheme that allows for selection effects. This means we usually need to compare distributions that are consistent with a broad range of possible observations, with actual experimental data. One controversial means of making this comparison is by invoking the 'principle of mediocrity': that is, (...) the principle that we are typical of the reference class implicit in the conjunction of the theory and the conditionalization scheme. In this paper, I quantitatively assess the principle of mediocrity in a range of cosmological settings, employing 'xerographic distributions' to impose a variety of assumptions regarding typicality. I find that for a fixed theory, the assumption that we are typical gives rise to higher likelihoods for our observations. If, however, one allows both the underlying theory and the assumption of typicality to vary, then the assumption of typicality does not always provide the highest likelihoods. Interpreted from a Bayesian perspective, these results support the claim that when one has the freedom to consider different combinations of theories and xerographic distributions, one should favor the framework that has the highest posterior probability; and then from this framework one can infer, in particular, how typical we are. In this way, the invocation of the principle of mediocrity is more questionable than has been recently claimed. (shrink)

An approach to testing theories describing a multiverse, that has gained interest of late, involves comparing theory-generated probability distributions over observables with their experimentally measured values. It is likely that such distributions, were we indeed able to calculate them unambiguously, will assign low probabilities to any such experimental measurements. An alternative to thereby rejecting these theories, is to conditionalize the distributions involved by restricting attention to domains of the multiverse in which we might arise. In order to elicit a crisp (...) prediction, however, one needs to make a further assumption about how typical we are of the chosen domains. In this paper, we investigate interactions between the spectra of available assumptions regarding both conditionalization and typicality, and draw out the effects of these interactions in a concrete setting; namely, on predictions of the total number of species that contribute significantly to dark matter. In particular, for each conditionalization scheme studied, we analyze how correlations between densities of different dark matter species affect the prediction, and explicate the effects of assumptions regarding typicality. We find that the effects of correlations can depend on the conditionalization scheme, and that in each case atypicality can significantly change the prediction. In doing so, we demonstrate the existence of overlaps in the predictions of different "frameworks" consisting of conjunctions of theory, conditionalization scheme and typicality assumption. This conclusion highlights the acute challenges involved in using such tests to identify a preferred framework that aims to describe our observational situation in a multiverse. (shrink)

We introduce a mathematical framework for quantifying fine-tuning in general physical settings. In particular, we identify two distinct perspectives on fine-tuning, namely, a local and a global perspective --- and develop corresponding measures. These measures apply broadly to settings characterized by an arbitrary number of observables whose values are dependent on an arbitrary number of parameters. We illustrate our formalism by quantifying fine-tuning as it arises in two pertinent astrophysical settings: in models where a significant fraction of the dark matter (...) in the universe is in the form of primordial black holes, and in scenarios that derive the fraction of protons in habitable dark-matter halos from underlying models of cosmic inflation. (shrink)

In Defeating Dr. Evil with Self-Locating Belief, Adam Elga proposes and defends a principle of indifference for self-locating beliefs: if an individual is confident that his world contains more than one individual who is in a state subjectively indistinguishable from his own, then he should assign equal credences to the hypotheses that he is any one of these individuals. Through a sequence of thought experiments, Elga in effect claims that he can derive the credence function that should apply in such (...) situations, thus justifying his principle of indifference. Here we argue, using a Bayesian approach, that Elga’s reasoning is circular: in analyzing the third of his thought experiments, he uses an assertion that is justifiable only if one assumes, from the start, the principle of indifference that he is attempting to justify. We agree with Elga that the assumption of equal credences is a very reasonable principle, in the absence of any reason to assign unequal credences, but we do not agree that the equality of credences can be so derived. (shrink)

We analyze the flow into inflation for generic "single-clock" systems, by combining an effective field theory approach with a dynamical-systems analysis. In this approach, we construct an expansion for the potential-like term in the effective action as a function of time, rather than specifying a particular functional dependence on a scalar field. We may then identify fixed points in the effective phase space for such systems, order-by-order, as various constraints are placed on the Mth time derivative of the potential-like function. (...) For relatively simple systems, we find significant probability for the background spacetime to flow into an inflationary state, and for inflation to persist for at least 60 efolds. Moreover, for systems that are compatible with single-scalar-field realizations, we find a single, universal functional form for the effective potential, $V$, which is similar to the well-studied potential for power-law inflation. We discuss the compatibility of such dynamical systems with observational constraints. (shrink)

Localizing content in neural networks provides a bridge to understanding the way in which the brain stores and processes information. In this paper, I propose the existence of polytopes in the state space of the hidden layer of feedforward neural networks as vehicles of content. I analyze these geometrical structures from an information-theoretic point of view, invoking mutual information to help define the content stored within them. I establish how this proposal addresses the problem of misclassification and provide a novel (...) solution to the disjunction problem, which hinges on the precise nature of the causal-informational framework for content advocated herein. (shrink)

Extracting predictions from cosmological theories that describe a multiverse, for what we are likely to observe in our domain, is crucial to establishing the validity of these theories. One way to extract such predictions is from theory-generated probability distributions that allow for selection effects---generally expressed in terms of assumptions about anthropic conditionalization and how typical we are. In this paper, I urge three lessons about typicality in multiverse settings. Because it is difficult to characterize our observational situation in the multiverse, (...) we cannot assume that we are typical : nor can we ignore the issue of typicality, for it has a measurable impact on predictions for our observations. There are spectra of assumptions about both conditionalization and typicality, which lead to coincident predictions for our observations, leading to problems of confirmation in multiverse cosmology. And moreover, when one has the freedom to consider competing theories of the multiverse, the assumption of typicality may not lead to the highest likelihoods for our observations. These three entwined aspects of typicality imply that positive assertions about our typicality, such as the 'principle of mediocrity', are more questionable than has been recently claimed. (shrink)