Research Papers

Over the last two years, The Alan Turing Institute and the Information Commissioner's Office (ICO) have been working together to discover ways to tackle the difficult issues surrounding explainable AI. The ultimate product of this joint endeavour, Explaining decisions made with AI, published in May 2020, is the most comprehensive practical guidance on AI explanation produced anywhere to date. We have put together this workbook to help support the uptake of that guidance. The goal of the workbook is to summarise some of main themes from Explaining decisions made with AI and then to provide the materials for a workshop exercise that has been built around a use case created to help you gain a flavour of how to put the guidance into practice. In the first three sections, we run through the basics of Explaining decisions made with AI. We provide a precis of the four principles of AI explainability, the typology of AI explanations, and the tasks involved in the explanation-aware design, development, and use of AI/ML systems. We then provide some reflection questions, which are intended to be a launching pad for group discussion, and a starting point for the case-study-based exercise that we have included as Appendix B. In Appendix A, we go into more detailed suggestions about how to organise the workshop. These recommendations are based on two workshops we had the privilege of co-hosting with our colleagues from the ICO and Manchester Metropolitan University in January 2021. The participants of these workshops came from both the private and public sectors, and we are extremely grateful to them for their energy, enthusiasm, and tremendous insight. This workbook would simply not exist without the commitment and keenness of all our collaborators and workshop participants.

A local surrogate for an AI-model correcting a simpler 'base' model is introduced representing an analytical method to yield explanations of AI-predictions. The approach is studied here in the context of the base model being linear regression. The AI-model approximates the residual error of the linear model and the explanations are formulated in terms of the change of the interpretable base model's parameters. Criteria are formulated for the precise relation between lost accuracy of the surrogate, the accuracy of the AI-model, and the surrogate fidelity. It is shown that, assuming a certain maximal amount of noise in the observed data, these criteria induce neighborhoods of the instances to be explained which have an ideal size in terms of maximal accuracy and fidelity.

With recent progress in the field of Explainable Artificial Intelligence (XAI) and increasing use in practice, the need for an evaluation of different XAI methods and their explanation quality in practical usage scenarios arises. For this purpose, we present VitrAI, which is a web-based service with the goal of uniformly demonstrating four different XAI algorithms in the context of three real life scenarios and evaluating their performance and comprehensibility for humans. This work reveals practical obstacles when adopting XAI methods and gives qualitative estimates on how well different approaches perform in said scenarios.

Conditional Variational Auto Encoders (VAE) are gathering significant attention as an Explainable Artificial Intelligence (XAI) tool. The codes in the latent space provide a theoretically sound way to produce counterfactuals, i.e. alterations resulting from an intervention on a targeted semantic feature. To be applied on real images more complex models are needed, such as Hierarchical CVAE. This comes with a challenge as the naive conditioning is no longer effective. In this paper we show how relaxing the effect of the posterior leads to successful counterfactuals and we introduce VAEX an Hierarchical VAE designed for this approach that can visually audit a classifier in applications.

Chess as a Testing Grounds for the Oracle Approach to AI Safety

James D. Miller, Roman Yampolskiy, Olle Haggstrom +1

To reduce the danger of powerful super-intelligent AIs, we might make the first such AIs oracles that can only send and receive messages. This paper proposes a possibly practical means of using machine learning to create two classes of narrow AI oracles that would provide chess advice: those aligned with the player's interest, and those that want the player to lose and give deceptively bad advice. The player would be uncertain which type of oracle it was interacting with. As the oracles would be vastly more intelligent than the player in the domain of chess, experience with these oracles might help us prepare for future artificial general intelligence oracles.

For feature selection and related problems, we introduce the notion of classification game, a cooperative game, with features as players and hinge loss based characteristic function and relate a feature's contribution to Shapley value based error apportioning (SVEA) of total training error. Our major contribution is ($\star$) to show that for any dataset the threshold 0 on SVEA value identifies feature subset whose joint interactions for label prediction is significant or those features that span a subspace where the data is predominantly lying. In addition, our scheme ($\star$) identifies the features on which Bayes classifier doesn't depend but any surrogate loss function based finite sample classifier does; this contributes to the excess $0$-$1$ risk of such a classifier, ($\star$) estimates unknown true hinge risk of a feature, and ($\star$) relate the stability property of an allocation and negative valued SVEA by designing the analogue of core of classification game. Due to Shapley value's computationally expensive nature, we build on a known Monte Carlo based approximation algorithm that computes characteristic function (Linear Programs) only when needed. We address the potential sample bias problem in feature selection by providing interval estimates for SVEA values obtained from multiple sub-samples. We illustrate all the above aspects on various synthetic and real datasets and show that our scheme achieves better results than existing recursive feature elimination technique and ReliefF in most cases. Our theoretically grounded classification game in terms of well defined characteristic function offers interpretability (which we formalize in terms of final task) and explainability of our framework, including identification of important features.

European Law now requires AI to be explainable in the context of adverse decisions affecting European Union (EU) citizens. At the same time, it is expected that there will be increasing instances of AI failure as it operates on imperfect data. This paper puts forward a neurally-inspired framework called decision stacks that can provide for a way forward in research aimed at developing explainable AI. Leveraging findings from memory systems in biological brains, the decision stack framework operationalizes the definition of explainability and then proposes a test that can potentially reveal how a given AI decision came to its conclusion.

The use of deep neural networks to make high risk decisions creates a need for global and local explanations so that users and experts have confidence in the modeling algorithms. We introduce a novel technique to find global and local explanations for time series data used in binary classification machine learning systems. We identify the most salient of the original features used by a black box model to distinguish between classes. The explanation can be made on categorical, continuous, and time series data and can be generalized to any binary classification model. The analysis is conducted on time series data to train a long short-term memory deep neural network and uses the time dependent structure of the underlying features in the explanation. The proposed technique attributes weights to features to explain an observations risk of belonging to a class as a multiplicative factor of a base hazard rate. We use a variation of the Cox Proportional Hazards regression, a Generalized Additive Model, to explain the effect of variables upon the probability of an in-class response for a score output from the black box model. The covariates incorporate time dependence structure in the features so the explanation is inclusive of the underlying time series data structure.

Saliency map generation techniques are at the forefront of explainable AI literature for a broad range of machine learning applications. Our goal is to question the limits of these approaches on more complex tasks. In this paper we apply Layer-Wise Relevance Propagation (LRP) to a sequence-to-sequence attention model trained on a text summarization dataset. We obtain unexpected saliency maps and discuss the rightfulness of these "explanations". We argue that we need a quantitative way of testing the counterfactual case to judge the truthfulness of the saliency maps. We suggest a protocol to check the validity of the importance attributed to the input and show that the saliency maps obtained sometimes capture the real use of the input features by the network, and sometimes do not. We use this example to discuss how careful we need to be when accepting them as explanation.

Interpretation of Feature Space using Multi-Channel Attentional Sub-Networks

Masanari Kimura, Masayuki Tanaka

Convolutional Neural Networks have achieved impressive results in various tasks, but interpreting the internal mechanism is a challenging problem. To tackle this problem, we exploit a multi-channel attention mechanism in feature space. Our network architecture allows us to obtain an attention mask for each feature while existing CNN visualization methods provide only a common attention mask for all features. We apply the proposed multi-channel attention mechanism to multi-attribute recognition task. We can obtain different attention mask for each feature and for each attribute. Those analyses give us deeper insight into the feature space of CNNs. The experimental results for the benchmark dataset show that the proposed method gives high interpretability to humans while accurately grasping the attributes of the data.

Converting malware into images followed by vision-based deep learning algorithms has shown superior threat detection efficacy compared with classical machine learning algorithms. When malware are visualized as images, visual-based interpretation schemes can also be applied to extract insights of why individual samples are classified as malicious. In this work, via two case studies of dynamic malware classification, we extend the local interpretable model-agnostic explanation algorithm to explain image-based dynamic malware classification and examine its interpretation fidelity. For both case studies, we first train deep learning models via transfer learning on malware images, demonstrate high classification effectiveness, apply an explanation method on the images, and correlate the results back to the samples to validate whether the algorithmic insights are consistent with security domain expertise. In our first case study, the interpretation framework identifies indirect calls that uniquely characterize the underlying exploit behavior of a malware family. In our second case study, the interpretation framework extracts insightful information such as cryptography-related APIs when applied on images created from API existence, but generate ambiguous interpretation on images created from API sequences and frequencies. Our findings indicate that current image-based interpretation techniques are promising for explaining vision-based malware classification. We continue to develop image-based interpretation schemes specifically for security applications.

With the success of deep learning, recent efforts have been focused on analyzing how learned networks make their classifications. We are interested in analyzing the network output based on the network structure and information flow through the network layers. We contribute an algorithm for 1) analyzing a deep network to find neurons that are 'important' in terms of the network classification outcome, and 2)automatically labeling the patches of the input image that activate these important neurons. We propose several measures of importance for neurons and demonstrate that our technique can be used to gain insight into, and explain how a network decomposes an image to make its final classification.