arXiv subjects

Jonathan G. Richens

Publications and source records attributed to Jonathan G. Richens.

7 recordsLinked to original sources

arXiv2022

Counterfactual harm

To act safely and ethically in the real world, agents must be able to reason about harm and avoid harmful actions. However, to date there is no statistical method for measuring harm and factoring it into algorithmic decisions. In this paper we propose the first formal definition of harm and benefit using causal models. We show that any factual definition of harm must violate basic intuitions in certain scenarios, and show that standard machine learning algorithms that cannot perform counterfactual reasoning are guaranteed to pursue harmful policies following distributional shifts. We use our definition of harm to devise a framework for harm-averse decision making using counterfactual objective functions. We demonstrate this framework on the problem of identifying optimal drug doses using a dose-response model learned from randomized control trial data. We find that the standard method of selecting doses using treatment effects results in unnecessarily harmful doses, while our counterfactual approach allows us to identify doses that are significantly less harmful without sacrificing efficacy.

cs.AI↗

arXiv2021

Leveraging directed causal discovery to detect latent common causes

The discovery of causal relationships is a fundamental problem in science and medicine. In recent years, many elegant approaches to discovering causal relationships between two variables from observational data have been proposed. However, most of these deal only with purely directed causal relationships and cannot detect latent common causes. Here, we devise a general heuristic which takes a causal discovery algorithm that can only distinguish purely directed causal relations and modifies it to also detect latent common causes. We apply our method to two directed causal discovery algorithms, the Information Geometric Causal Inference of (Daniusis et al., 2010) and the Kernel Conditional Deviance for Causal Inference of (Mitrovic, Sejdinovic, & Teh, 2018), and extensively test on synthetic data -- detecting latent common causes in additive, multiplicative and complex noise regimes -- and on real data, where we are able to detect known common causes. In addition to detecting latent common causes, our experiments demonstrate that both the modified algorithms preserve the performance of the original in distinguishing directed causal relations.

stat.ML↗

arXiv2020

Counterfactual diagnosis

Machine learning promises to revolutionize clinical decision making and diagnosis. In medical diagnosis a doctor aims to explain a patient's symptoms by determining the diseases \emph{causing} them. However, existing diagnostic algorithms are purely associative, identifying diseases that are strongly correlated with a patients symptoms and medical history. We show that this inability to disentangle correlation from causation can result in sub-optimal or dangerous diagnoses. To overcome this, we reformulate diagnosis as a counterfactual inference task and derive new counterfactual diagnostic algorithms. We show that this approach is closer to the diagnostic reasoning of clinicians and significantly improves the accuracy and safety of the resulting diagnoses. We compare our counterfactual algorithm to the standard Bayesian diagnostic algorithm and a cohort of 44 doctors using a test set of clinical vignettes. While the Bayesian algorithm achieves an accuracy comparable to the average doctor, placing in the top 48% of doctors in our cohort, our counterfactual algorithm places in the top 25% of doctors, achieving expert clinical accuracy. This improvement is achieved simply by changing how we query our model, without requiring any additional model improvements. Our results show that counterfactual reasoning is a vital missing ingredient for applying machine learning to medical diagnosis.

stat.ML↗

arXiv2020

MultiVerse: Causal Reasoning using Importance Sampling in Probabilistic Programming

We elaborate on using importance sampling for causal reasoning, in particular for counterfactual inference. We show how this can be implemented natively in probabilistic programming. By considering the structure of the counterfactual query, one can significantly optimise the inference process. We also consider design choices to enable further optimisations. We introduce MultiVerse, a probabilistic programming prototype engine for approximate causal reasoning. We provide experimental results and compare with Pyro, an existing probabilistic programming framework with some of causal reasoning tools.

cs.AI↗

arXiv2017

From single-shot to general work extraction with bounded fluctuations in work

In the standard framework of thermodynamics the work produced or consumed in a process is a random variable whose average value is bounded by the change in the free energy of the system. This work is calculated without regard for the size of its fluctuations. We find that in some processes, such as reversible cooling, the fluctuations of the work can diverge. Small or fragile thermal machines may be unable to cope with large fluctuations. Hence, with the present focus on nano- scale thermodynamics, we analyse how thermodynamic efficiency rates are modified when the size of the fluctuations around the average is restricted. We quantify the work content and work of formation of any state when the work fluctuations are bounded by a given amount c. By varying c we interpolate between the standard free energy c = infinity and the min-free energy c = 0, defined in the context of single-shot thermodynamics. We derive fundamental relations between average work and its fluctuations, and explore the emergence of irreversibility and partial order on state transformations when bounding fluctuations. We also study the efficiency a single qubit thermal engine model with constrained fluctuations, and derive the corrections to the Carnot efficiency.

quant-ph↗

arXiv2017

Finite-bath corrections to the second law of thermodynamics

The second law of thermodynamics states that a system in contact with a heat bath can undergo a transformation if and only if its free energy decreases. However, the "if" part of this statement is only true when the effective heat bath is infinite. In this article we remove this idealization and derive corrections to the second law in the case where the bath has a finite size, or equivalently finite heat capacity. This can also be translated to processes lasting a finite time, and we show that thermodynamical reversibility is lost in this regime. We do so in full generality, that is without assuming any particular model for the bath, the only parameters defining the bath are its temperature and heat capacity. We find connections with second order Shannon information theory, in particular in the case of Landauer erasure. We also consider the case of non-fluctuating work, and derive finite-bath corrections to the min and max free energies employed in single-shot thermodynamics.

quant-ph↗

arXiv2016

Entanglement is an inevitable feature of any non-classical theory

One of the most striking features of quantum theory is the existence of entangled states, responsible for Einstein's so called "spooky action at a distance". These states emerge from the mathematical formalism of quantum theory, but to date we do not have a clear idea of which physical principles give rise to entanglement. Why does quantum theory have entangled states? Would any theory superseding classical theory have entangled states, or is quantum theory special? We demonstrate that without entanglement, non-classical degrees of freedom cannot reversibly interact. We present two postulates, no-cloning / no-broadcasting and local transitivity, either of which are sufficient to imply the existence of entangled states in any non-classical theory with reversible interactions. Therefore we argue that entanglement is an inevitable feature of a non-classical universe.

quant-ph↗