Christopher G. Lucas

Balint Gyevnar, Cheng Wang, Christopher G. Lucas et al.

We present CEMA: Causal Explanations in Multi-Agent systems; a framework for creating causal natural language explanations of an agent's decisions in dynamic sequential multi-agent systems to build more trustworthy autonomous agents. Unlike prior work that assumes a fixed causal structure, CEMA only requires a probabilistic model for forward-simulating the state of the system. Using such a model, CEMA simulates counterfactual worlds that identify the salient causes behind the agent's decisions. We evaluate CEMA on the task of motion planning for autonomous driving and test it in diverse simulated scenarios. We show that CEMA correctly and robustly identifies the causes behind the agent's decisions, even when a large number of other agents is present, and show via a user study that CEMA's explanations have a positive effect on participants' trust in autonomous vehicles and are rated as high as high-quality baseline explanations elicited from other participants. We release the collected explanations with annotations as the HEADD dataset.

Chentian Jiang, Christopher G. Lucas

How do people actively learn to learn? That is, how and when do people choose actions that facilitate long-term learning and choosing future actions that are more informative? We explore these questions in the domain of active causal learning. We propose a hierarchical Bayesian model that goes beyond past models by predicting that people pursue information not only about the causal relationship at hand but also about causal overhypotheses$\unicode{x2014}$abstract beliefs about causal relationships that span multiple situations and constrain how we learn the specifics in each situation. In two active "blicket detector" experiments with 14 between-subjects manipulations, our model was supported by both qualitative trends in participant behavior and an individual-differences-based model comparison. Our results suggest when there are abstract similarities across active causal learning problems, people readily learn and transfer overhypotheses about these similarities. Moreover, people exploit these overhypotheses to facilitate long-term active learning.

Alessandro B. Palmarini, Christopher G. Lucas, N. Siddharth

DreamCoder is an inductive program synthesis system that, whilst solving problems, learns to simplify search in an iterative wake-sleep procedure. The cost of search is amortized by training a neural search policy, reducing search breadth and effectively "compiling" useful information to compose program solutions across tasks. Additionally, a library of program components is learnt to compress and express discovered solutions in fewer components, reducing search depth. We present a novel approach for library learning that directly leverages the neural search policy, effectively "decompiling" its amortized knowledge to extract relevant program components. This provides stronger amortized inference: the amortized knowledge learnt to reduce search breadth is now also used to reduce search depth. We integrate our approach with DreamCoder and demonstrate faster domain proficiency with improved generalization on a range of domains, particularly when fewer example solutions are available.

Sigrid Passano Hellan, Christopher G. Lucas, Nigel H. Goddard

Bayesian optimisation is a powerful method for optimising black-box functions, popular in settings where the true function is expensive to evaluate and no gradient information is available. Bayesian optimisation can improve responses to many optimisation problems within climate change for which simulator models are unavailable or expensive to sample from. While there have been several demonstrations of climate-related applications, there has been no unifying review of applications and benchmarks. We provide such a review here, to encourage the use of Bayesian optimisation for important and well-suited applications. We identify four main application domains: material discovery, wind farm layout, optimal renewable control and environmental monitoring. For each domain we identify a public benchmark or data set that is easy to use and evaluate systems against, while being representative of real-world problems. Due to the lack of a suitable benchmark for environmental monitoring, we propose LAQN-BO, based on air pollution data. Our contributions are: a) summarising Bayesian optimisation applications related to climate change; b) identifying a representative range of benchmarks, providing example code where necessary; and c) introducing a new benchmark, LAQN-BO.

Sigrid Passano Hellan, Christopher G. Lucas, Nigel H. Goddard

Transfer learning for Bayesian optimisation has generally assumed a strong similarity between optimisation tasks, with at least a subset having similar optimal inputs. This assumption can reduce computational costs, but it is violated in a wide range of optimisation problems where transfer learning may nonetheless be useful. We replace this assumption with a weaker one only requiring the shape of the optimisation landscape to be similar, and analyse the recent method Prior Learning for Bayesian Optimisation - PLeBO - in this setting. By learning priors for the hyperparameters of the Gaussian process surrogate model we can better approximate the underlying function, especially for few function evaluations. We validate the learned priors and compare to a breadth of transfer learning approaches, using synthetic data and a recent air pollution optimisation problem as benchmarks. We show that PLeBO and prior transfer find good inputs in fewer evaluations.

Samuel Garcin, James Doran, Shangmin Guo et al.

Autonomous agents trained using deep reinforcement learning (RL) often lack the ability to successfully generalise to new environments, even when these environments share characteristics with the ones they have encountered during training. In this work, we investigate how the sampling of individual environment instances, or levels, affects the zero-shot generalisation (ZSG) ability of RL agents. We discover that, for deep actor-critic architectures sharing their base layers, prioritising levels according to their value loss minimises the mutual information between the agent's internal representation and the set of training levels in the generated training data. This provides a novel theoretical justification for the regularisation achieved by certain adaptive sampling strategies. We then turn our attention to unsupervised environment design (UED) methods, which assume control over level generation. We find that existing UED methods can significantly shift the training distribution, which translates to low ZSG performance. To prevent both overfitting and distributional shift, we introduce data-regularised environment design (DRED). DRED generates levels using a generative model trained to approximate the ground truth distribution of an initial set of level parameters. Through its grounding, DRED achieves significant improvements in ZSG over adaptive level sampling strategies and UED methods. Our code and experimental data are available at https://github.com/uoe-agents/dred.

Samuel Garcin, James Doran, Shangmin Guo et al.

A key limitation preventing the wider adoption of autonomous agents trained via deep reinforcement learning (RL) is their limited ability to generalise to new environments, even when these share similar characteristics with environments encountered during training. In this work, we investigate how a non-uniform sampling strategy of individual environment instances, or levels, affects the zero-shot generalisation (ZSG) ability of RL agents, considering two failure modes: overfitting and over-generalisation. As a first step, we measure the mutual information (MI) between the agent's internal representation and the set of training levels, which we find to be well-correlated to instance overfitting. In contrast to uniform sampling, adaptive sampling strategies prioritising levels based on their value loss are more effective at maintaining lower MI, which provides a novel theoretical justification for this class of techniques. We then turn our attention to unsupervised environment design (UED) methods, which adaptively generate new training levels and minimise MI more effectively than methods sampling from a fixed set. However, we find UED methods significantly shift the training distribution, resulting in over-generalisation and worse ZSG performance over the distribution of interest. To prevent both instance overfitting and over-generalisation, we introduce self-supervised environment design (SSED). SSED generates levels using a variational autoencoder, effectively reducing MI while minimising the shift with the distribution of interest, and leads to statistically significant improvements in ZSG over fixed-set level sampling strategies and UED methods.

Max Taylor-Davies, Christopher G. Lucas

To successfully navigate its environment, an agent must construct and maintain representations of the other agents that it encounters. Such representations are useful for many tasks, but they are not without cost. As a result, agents must make decisions regarding how much information they choose to store about the agents in their environment. Using selective social learning as an example task, we motivate the problem of finding agent representations that optimally trade off between downstream utility and information cost, and illustrate two example approaches to resource-constrained social representation.

Elizabeth Mieczkowski, Ruaridh Mon-Williams, Neil Bramley et al.

When should we encourage specialization in multi-agent systems versus train generalists that perform the entire task independently? We propose that specialization largely depends on task parallelizability: the potential for multiple agents to execute task components concurrently. Drawing inspiration from Amdahl's Law in distributed systems, we present a closed-form bound that predicts when specialization improves performance, depending only on task concurrency and team size. We validate our model on two standard MARL benchmarks that represent opposite regimes -- StarCraft Multi-Agent Challenge (SMAC, unlimited concurrency) and Multi-Particle Environment (MPE, unit-capacity bottlenecks) -- and observe close alignment between the bound at each extreme and an empirical measure of specialization. Three follow-up experiments in Overcooked-AI demonstrate that the model works in environments with more complex spatial and resource bottlenecks that allow for a range of strategies. Beyond prediction, the bound also serves as a diagnostic tool, highlighting biases in MARL training algorithms that cause sub-optimal convergence to specialist strategies with larger state spaces.

Simon Valentin, Steven Kleinegesse, Neil R. Bramley et al.

Computational models are powerful tools for understanding human cognition and behavior. They let us express our theories clearly and precisely, and offer predictions that can be subtle and often counter-intuitive. However, this same richness and ability to surprise means our scientific intuitions and traditional tools are ill-suited to designing experiments to test and compare these models. To avoid these pitfalls and realize the full potential of computational modeling, we require tools to design experiments that provide clear answers about what models explain human behavior and the auxiliary assumptions those models must make. Bayesian optimal experimental design (BOED) formalizes the search for optimal experimental designs by identifying experiments that are expected to yield informative data. In this work, we provide a tutorial on leveraging recent advances in BOED and machine learning to find optimal experiments for any kind of model that we can simulate data from, and show how by-products of this procedure allow for quick and straightforward evaluation of models and their parameters against real experimental data. As a case study, we consider theories of how people balance exploration and exploitation in multi-armed bandit decision-making tasks. We validate the presented approach using simulations and a real-world experiment. As compared to experimental designs commonly used in the literature, we show that our optimal designs more efficiently determine which of a set of models best account for individual human behavior, and more efficiently characterize behavior given a preferred model. At the same time, formalizing a scientific question such that it can be adequately addressed with BOED can be challenging and we discuss several potential caveats and pitfalls that practitioners should be aware of. We provide code and tutorial notebooks to replicate all analyses.

Sigrid Passano Hellan, Christopher G. Lucas, Nigel H. Goddard

Air pollution is one of the leading causes of mortality globally, resulting in millions of deaths each year. Efficient monitoring is important to measure exposure and enforce legal limits. New low-cost sensors can be deployed in greater numbers and in more varied locations, motivating the problem of efficient automated placement. Previous work suggests Bayesian optimisation is an appropriate method, but only considered a satellite data set, with data aggregated over all altitudes. It is ground-level pollution, that humans breathe, which matters most. We improve on those results using hierarchical models and evaluate our models on urban pollution data in London to show that Bayesian optimisation can be successfully applied to the problem.

Bonan Zhao, Christopher G. Lucas, Neil R. Bramley

We present a novel task that measures how people generalize objects' causal powers based on observing a single (Experiment 1) or a few (Experiment 2) causal interactions between object pairs. We propose a computational modeling framework that can synthesize human-like generalization patterns in our task setting, and sheds light on how people may navigate the compositional space of possible causal functions and categories efficiently. Our modeling framework combines a causal function generator that makes use of agent and recipient objects' features and relations, and a Bayesian non-parametric inference process to govern the degree of similarity-based generalization. Our model has a natural "resource-rational" variant that outperforms a naive Bayesian account in describing participants, in particular reproducing a generalization-order effect and causal asymmetry observed in our behavioral experiments. We argue that this modeling framework provides a computationally plausible mechanism for real world causal generalization.

Simon Valentin, Steven Kleinegesse, Neil R. Bramley et al.

Bayesian optimal experimental design (BOED) is a methodology to identify experiments that are expected to yield informative data. Recent work in cognitive science considered BOED for computational models of human behavior with tractable and known likelihood functions. However, tractability often comes at the cost of realism; simulator models that can capture the richness of human behavior are often intractable. In this work, we combine recent advances in BOED and approximate inference for intractable models, using machine-learning methods to find optimal experimental designs, approximate sufficient summary statistics and amortized posterior distributions. Our simulation experiments on multi-armed bandit tasks show that our method results in improved model discrimination and parameter estimation, as compared to experimental designs commonly used in the literature.

Sigrid Passano Hellan, Christopher G. Lucas, Nigel H. Goddard

Air pollution is one of the most important causes of mortality in the world. Monitoring air pollution is useful to learn more about the link between health and pollutants, and to identify areas for intervention. Such monitoring is expensive, so it is important to place sensors as efficiently as possible. Bayesian optimisation has proven useful in choosing sensor locations, but typically relies on kernel functions that neglect the statistical structure of air pollution, such as the tendency of pollution to propagate in the prevailing wind direction. We describe two new wind-informed kernels and investigate their advantage for the task of actively learning locations of maximum pollution using Bayesian optimisation.

Andrew Gordon Wilson, Christoph Dann, Christopher G. Lucas et al.

Bayesian nonparametric models, such as Gaussian processes, provide a compelling framework for automatic statistical modelling: these models have a high degree of flexibility, and automatically calibrated complexity. However, automating human expertise remains elusive; for example, Gaussian processes with standard kernels struggle on function extrapolation problems that are trivial for human learners. In this paper, we create function extrapolation problems and acquire human responses, and then design a kernel learning framework to reverse engineer the inductive biases of human learners across a set of behavioral experiments. We use the learned kernels to gain psychological insights and to extrapolate in human-like ways that go beyond traditional stationary and polynomial kernels. Finally, we investigate Occam's razor in human and Gaussian process based function learning.