There has been a significant research effort focused on explaining predictive models, for example through post-hoc explainability and recourse methods. Most of the proposed techniques operate upon a single, fixed, predictive model. However, it is well-known that given a dataset and a predictive task, there may be a multiplicity of models that solve the problem (nearly) equally well. In this work, we investigate the implications of this kind of model indeterminacy on the post-hoc explanations of predictive models. We show how it can lead to explanatory multiplicity, and we explore the underlying drivers. We show how predictive multiplicity, and the related concept of epistemic uncertainty, are not reliable indicators of explanatory multiplicity. We further illustrate how a set of models showing very similar aggregate performance on a test dataset may show large variations in their local explanations, i.e., for a specific input. We explore these effects for Shapley value based explanations on three risk assessment datasets. Our results indicate that model indeterminacy may have a substantial impact on explanations in practice, leading to inconsistent and even contradicting explanations.

In this work, we focus on the use of influence functions to identify relevant training examples that one might hope “explain” the predictions of a machine learning model. One shortcoming of influence functions is that the training examples deemed most “influential” are often outliers or mislabelled, making them poor choices for explanation. In order to address this shortcoming, we separate the role of global versus local influence. We introduce RelatIF, a new class of criteria for choosing relevant training examples by way of an optimization objective that places a constraint on global influence. RelatIF considers the local influence that an explanatory example has on a prediction relative to its global effects on the model. In empirical evaluations, we find that the examples returned by RelatIF are more intuitive when compared to those found using influence functions.

Popular word embedding algorithms exhibit stereotypical biases, such as gender bias. The widespread use of these algorithms in machine learning systems can amplify stereotypes in important contexts. Although some methods have been developed to mitigate this problem, how word embedding biases arise during training is poorly understood. In this work we develop a technique to address this question. Given a word embedding, our method reveals how perturbing the training corpus would affect the resulting embedding bias. By tracing the origins of word embedding bias back to the original training documents, one can identify subsets of documents whose removal would most reduce bias. We demonstrate our methodology on Wikipedia and New York Times corpora, and find it to be very accurate.

Demand-responsive transport (DRT) systems provide flexible transport services for passengers who request door-to-door rides in shared-ride mode without fixed routes and schedules. DRT systems face interesting coordination challenges. For example, one has to design cost-sharing mechanisms for offering fare quotes to potential passengers so that all passengers are treated fairly. The main issue is how the operating costs of the DRT system should be shared among the passengers (given that different passengers cause different amounts of inconvenience to the other passengers), taking into account that DRT systems should provide fare quotes instantaneously without knowing future ride request submissions. We determine properties of cost-sharing mechanisms that make DRT systems attractive to both the transport providers and passengers, namely online fairness, immediate response, individual rationality, budget balance, and ex-post incentive compatibility. We propose a novel cost-sharing mechanism, which is called Proportional Online Cost Sharing (POCS), which provides passengers with upper bounds on their fares immediately after their ride request submissions despite missing knowledge of future ride request submissions, allowing them to accept their fare quotes or drop out. We examine how POCS satisfies these properties in theory and computational experiments.