The ability to substitute some resource or tool for another is a common and important human ability. For example, in cooking, we often lack an ingredient for a recipe and we solve this problem by finding a substitute ingredient. There are various ways that we may reason about this. Often we need to draw on commonsense reasoning to find a substitute. For instance, we can think of the properties of the missing item, and try to find similar items with similar properties. Despite the importance of substitution in human intelligence, there is a lack of a theoretical understanding of the faculty. To address this shortcoming, we propose a commonsense reasoning framework for conceptualizing and harnessing substitution. In order to ground our proposal, we focus on cooking. Though we believe the proposal can be straightforwardly adapted to other applications that require formalization of substitution. Our approach is to produce a general framework based on distance measures for determining similarity (e.g. between ingredients, or between processing steps), and on identifying inconsistencies between the logical representation of recipes and integrity constraints that we use to flag the need for mitigation (e.g. after substituting one kind of pasta for another in a recipe, we may identify an inconsistency in the cooking time, and this is resolved by updating the cooking time).

Given that AI systems are set to play a pivotal role in future decision-making processes, their trustworthiness and reliability are of critical concern. Due to their scale and complexity, modern AI systems resist direct interpretation, and alternative ways are needed to establish trust in those systems, and determine how well they align with human values. We argue that good measures of the information processing similarities between AI and humans, may be able to achieve these same ends. While Representational alignment (RA) approaches measure similarity between the internal states of two systems, the associated data can be expensive and difficult to collect for human systems. In contrast, Behavioural alignment (BA) comparisons are cheaper and easier, but questions remain as to their sensitivity and reliability. We propose two new behavioural alignment metrics misclassification agreement which measures the similarity between the errors of two systems on the same instances, and class-level error similarity which measures the similarity between the error distributions of two systems. We show that our metrics correlate well with RA metrics, and provide complementary information to another BA metric, within a range of domains, and set the scene for a new approach to value alignment.

Commonsense reasoning is a key aspect of human intelligence. If we are to develop robust and deep intelligent systems, then we need to understand the diversity and complexity of commonsense reasoning across the gamut of human activities. An interesting class of commonsense reasoning problems arises when people are faced with natural disasters. To investigate this topic, we present RESPONSE, a human-curated dataset containing 1789 annotated instances featuring 6037 sets of questions designed to assess LLMs’ commonsense reasoning in disaster situations across different time frames. The dataset includes problem descriptions, missing resources, time-sensitive solutions, and their justifications, with a subset validated by environmental engineers. Through both automatic metrics and human evaluation, we compare LLM-generated recommendations against human responses. Our findings show that even state-of-the-art models like GPT-4 achieve only 37% human-evaluated correctness for immediate response actions, highlighting significant room for improvement in LLMs’ ability for commonsense reasoning in crises.

Neural Disjunctive Normal Form (DNF) based models are powerful and interpretable approaches to neuro-symbolic learning and have shown promising results in classification and reinforcement learning settings without prior knowledge of the tasks. However, their performance is degraded by the thresholding of the post-training symbolic translation process. We show here that part of the performance degradation during translation is due to its failure to disentangle the learned knowledge represented in the form of the networks’ weights. We address this issue by proposing a new disentanglement method; by splitting nodes that encode nested rules into smaller independent nodes, we are able to better preserve the models’ performance. Through experiments on binary, multiclass, and multilabel classification tasks (including those requiring predicate invention), we demonstrate that our disentanglement method provides compact and interpretable logical representations for the neural DNF-based models, with performance closer to that of their pre-translation counterparts. Our code is available at https://github.com/kittykg/disentangling-ndnf-classification.

Probabilistic Event Calculus (PEC) is a logical framework for reasoning about actions and their effects in uncertain environments, which enables the representation of probabilistic narratives and computation of temporal projections. The PEC formalism offers significant advantages in interpretability and expressiveness for narrative reasoning. However, it lacks mechanisms for goal-directed reasoning. Our work bridges this gap by developing a formal translation of PEC domains into Markov Decision Processes (MDPs), introducing the concept of "action-taking situations" to preserve PEC’s flexible action semantics. The resulting PEC-MDP formalism enables the extensive collection of algorithms and theoretical tools developed for MDPs to be applied to PEC’s interpretable narrative domains. We demonstrate how the translation supports both temporal reasoning tasks and objective-driven planning, with methods for mapping learned policies back into human-readable PEC representations, maintaining interpretability while extending PEC’s capabilities.

We describe a general procedure for translating Epistemic Probabilistic Event Calculus (EPEC) action language domains into Answer Set Programs (ASP), and show how the Python-driven features of the ASP solver Clingo can be used to provide efficient computation in this probabilistic setting. EPEC supports probabilistic, epistemic reasoning in domains containing narratives that include both an agent's own action executions and environmentally triggered events. Some of the agent's actions may be belief-conditioned, and some may be imperfect sensing actions that alter the strengths of previously held beliefs. We show that our ASP implementation can be used to provide query answers that fully correspond to EPEC's own declarative, Bayesian-inspired semantics.

In the study of reasoning in neural networks, recent efforts have sought to improve consistency and coherence of sequence models, leading to important developments in the area of neuro-symbolic AI. In symbolic AI, the concepts of consistency and coherence can be defined and verified formally, but for neural networks these definitions are lacking. The provision of such formal definitions is crucial to offer a common basis for the quantitative evaluation and systematic comparison of connectionist, neuro-symbolic and transfer learning approaches. In this paper, we introduce formal definitions of consistency and coherence for neural systems. To illustrate the usefulness of our definitions, we propose a new dynamic relation-decoder model built around the principles of consistency and coherence. We compare our results with several existing relation-decoders using a partial transfer learning task based on a novel data set introduced in this paper. Our experiments show that relation-decoders that maintain consistency over unobserved regions of representation space retaincoherence across domains, whilst achieving better transfer learning performance.

Some online social networks (OSNs) allow users to define friendship-groups as reusable shortcuts for sharing information with multiple contacts. Posting exclusively to a friendship-group gives some privacy control, while supporting communication with (and within) this group. However, recipients of such posts may want to reuse content for their own social advantage, and can bypass existing controls by copy-pasting into a new post; this cross-posting poses privacy risks. This paper presents a learning to share approach that enables the incorporation of more nuanced privacy controls into OSNs. Specifically, we propose a reusable, adaptive software architecture that uses rigorous runtime analysis to help OSN users to make informed decisions about suitable audiences for their posts. This is achieved by supporting dynamic formation of recipient-groups that benefit social interactions while reducing privacy risks. We exemplify the use of our approach in the context of Facebook.

Probabilistic programming is an area of research that aims to develop general inference algorithms for probabilistic models expressed as probabilistic programs whose execution corresponds to inferring the parameters of those models. In this paper, we introduce a probabilistic programming language (PPL) based on abductive logic programming for performing inference in probabilistic models involving categorical distributions with Dirichlet priors. We encode these models as abductive logic programs enriched with probabilistic definitions and queries, and show how to execute and compile them to boolean formulas. Using the latter, we perform generalized inference using one of two proposed Markov Chain Monte Carlo (MCMC) sampling algorithms: an adaptation of uncollapsed Gibbs sampling from related work and a novel collapsed Gibbs sampling (CGS). We show that CGS converges faster than the uncollapsed version on a latent Dirichlet allocation (LDA) task using synthetic data. On similar data, we compare our PPL with LDA-specific algorithms and other PPLs. We find that all methods, except one, perform similarly and that the more expressive the PPL, the slower it is. We illustrate applications of our PPL on real data in two variants of LDA models (Seed and Cluster LDA), and in the repeated insertion model (RIM). In the latter, our PPL yields similar conclusions to inference with EM for Mallows models.

This paper addresses the making of security decisions, such as access-control decisions or spam filtering decisions, under uncertainty, when the benefit of doing so outweighs the need to absolutely guarantee these decisions are correct. For instance, when there are limited, costly, or failed communication channels to a policy-decision-point. Previously, local caching of decisions has been proposed, but when a correct decision is not available, either a policy-decision-point must be contacted, or a default decision used. We improve upon this model by using learned classifiers of access control decisions. These classifiers, trained on known decisions, infer decisions when an exact match has not been cached, and uses intuitive notions of utility, damage and uncertainty to determine when an inferred decision is preferred over contacting a remote PDP. Clearly there is uncertainty in the predicted decisions, introducing a degree of risk. Our solution proposes a mechanism to quantify the uncertainty of these decisions and allows administrators to bound the overall risk posture of the system. The learning component continuously refines its models based on inputs from a central policy server in cases where the risk is too high or there is too much uncertainty. We have validated our models by building a prototype system and evaluating it with requests from real access control policies. Our experiments show that over a range of system parameters, it is feasible to use machine learning methods to infer access control policies decisions. Thus our system yields several benefits, including reduced calls to the PDP, reducing latency and communication costs; increased net utility; and increased system survivability.

An understanding of information flow has many applications, including for maximizing marketing impact on social media, limiting malware propagation, and managing undesired disclosure of sensitive information. This paper presents scalable methods for both learning models of information flow in networks from data, based on the Independent Cascade Model, and predicting probabilities of unseen flow from these models. Our approach is based on a principled probabilistic construction and results compare favourably with existing methods in terms of accuracy of prediction and scalable evaluation, with the addition that we are able to evaluate a broader range of queries than previously shown, including probability of joint and/or conditional flow, as well as reflecting model uncertainty. Exact evaluation of flow probabilities is exponential in the number of edges and naive sampling can also be expensive, so we propose sampling in an efficient Markov-Chain Monte-Carlo fashion using the Metropolis-Hastings algorithm -- details described in the paper. We identify two types of data, those where the paths of past flows are known -- attributed data, and those where only the endpoints are known -- unattributed data. Both data types are addressed in this paper, including training methods, example real world data sets, and experimental evaluation. In particular, we investigate flow data from the Twitter microblogging service, exploring the flow of messages through retweets (tweet forwards) for the attributed case, and the propagation of hash tags (metadata tags) and urls for the unattributed case.

This paper concerns the resource management problem arising in public cycle sharing schemes, when some docking stations become empty and remain so while others fill to capacity. To alleviate this, managing companies move bicycles between docking stations in order to maximise the number of satisfied customers while minimising the movement cost. We identify Reinforcement learning (RL) as the most promising technique for finding good movement strategies in these networks, but conventional function-approximation RL methods do not scale well here, due to the quadratic growth in number of actions with network size. We propose the use of cooperating agents, namely Independent Learners, to partition the action space. To overcome the well known issue of coordination in Independent Learners, we combine a novel scheduling approach for asynchronous learning, with a modified Gradient-descent Sarsa(λ) algorithm to manage variable step-sizes. Our method competes with, and scales more favourably than, single-agent RL on a selection of simulated networks.

There has been increasing interest in automatic techniques for generating roles for role based access control, a process known as role mining. Most role mining approaches assume the input data is clean, and attempt to optimize the RBAC state. We examine role mining with noisy input data and suggest dividing the problem into two steps: noise removal and candidate role generation. We introduce an approach to use (non-binary) rank reduced matrix factorization to identify noise and experimentally show that it is effective at identifying noise in access control data. User- and permission-attributes can further be used to improve accuracy. Next, we show that our two-step approach is able to find candidate roles that are close to the roles mined from noise-less data. This method performs better than the approach of mining noisy data directly and offering the administrator increased control in the noise removal and candidate role generation phases. We note that our approach is applicable outside role engineering and may be used to identify errors or predict missing values in any access control matrix.