We present a system for parsing advice offered by a human to a machine. The advice is given in the form of conditional sentences in natural language, and the system generates a logic-based (machine-readable) rep-resentation of the advice, as appropriate for use by the machine in a downstream task. The system utilizes a “white-box” knowledge-based translation policy, which can be acquired iteratively in a developmental manner through a coaching process. We showcase this coaching process by demonstrating how linguistic annotations of sentences can be combined, through simple logic-based expressions, to carry out the translation task.

Abstract A Citizens’ assembly (CA) is a democratic innovation tool where a randomly selected group of citizens deliberate a topic over multiple rounds to generate, and then vote upon, policy recommendations. Despite growing popularity, little work exists on understanding how CA inputs, such as the expert selection process and the mixing method used for discussion groups, affect results. In this work, we model CA deliberation and opinion change as a multi-agent systems problem. We introduce and formalise a set of criteria for evaluating successful CAs using insight from previous CA trials and theoretical results. Although real-world trials meet these criteria, we show that finding a model that does so is non-trivial; through simulations and theoretical arguments, we show that established opinion change models fail at least one of these criteria. We therefore propose an augmented opinion change model with a latent ‘open-mindedness’ variable, which sufficiently captures people’s propensity to change opinion. We show that data from the CA of Scotland indicates a latent variable both exists and resembles the concept of open-mindedness in the literature. We calibrate parameters against real CA data, demonstrating our model’s ecological validity, before running simulations across a range of realistic global parameters, with each simulation satisfying our criteria. Specifically, simulations meet criteria regardless of expert selection, expert ordering, participant extremism, and sub-optimal participant grouping, which has ramifications for optimised algorithmic approaches in the computational CA space.

Natural target functions and tasks typically exhibit hierarchical modularity -- they can be broken down into simpler sub-functions that are organized in a hierarchy. Such sub-functions have two important features: they have a distinct set of inputs (input-separability) and they are reused as inputs higher in the hierarchy (reusability). Previous studies have established that hierarchically modular neural networks, which are inherently sparse, offer benefits such as learning efficiency, generalization, multi-task learning, and transfer. However, identifying the underlying sub-functions and their hierarchical structure for a given task can be challenging. The high-level question in this work is: if we learn a task using a sufficiently deep neural network, how can we uncover the underlying hierarchy of sub-functions in that task? As a starting point, we examine the domain of Boolean functions, where it is easier to determine whether a task is hierarchically modular. We propose an approach based on iterative unit and edge pruning (during training), combined with network analysis for module detection and hierarchy inference. Finally, we demonstrate that this method can uncover the hierarchical modularity of a wide range of Boolean functions and two vision tasks based on the MNIST digits dataset.

The unreflected promotion of diversity as a value in social interactions --- including in technology-mediated ones --- risks emphasizing the benefits of inclusion without recognizing the potential harm of failing to protect vulnerable individuals or account for the empowerment of marginalized groups. Adopting the position that technology is not value-neutral, we seek to answer the question of how technology-mediated social platforms can accommodate diversity by design by balancing the often tension-ridden principles of protection and inclusion. In this paper, we present our research program, developed strategy, as well as first analyses and results. Building on approaches from scenario analysis and Value Sensitive Design, we identify key arguments for a "diversity by design''-agenda. Furthermore, we discuss how these arguments can be operationalized and implemented in a diversity-aware chatbot and provide a critical reflection on the limits and drawbacks of the proposed approach.

We consider the problem of constructing rankings that exhibit a prescribed level of diversity. We take a black-box view of the ranking procedure itself, and choose, instead, to post-hoc curate any given ranking, deviating from the original ranking only to the extent allowed by any given constraints. The curation algorithm that we present is oblivious to how diversity is measured, and returns shuffled versions of the original rankings that are optimal in terms of their exhibited level of diversity. Our empirical evaluation on synthetic data demonstrates the effectiveness and efficiency of our methodology, across a number of diversity metrics from the literature.

Social networks constitute an important medium for social interaction where people communicate and formulate relationships in a way similar to what they do in real life. The analysis of the users' relationships in social networks can lead to new insights into human social behavior. Tie strength constitutes a core aspect of social relationships, which represents the importance of a relationship and the closeness of individuals. Understanding the key features of tie strength in social networks can assist in formulating more efficient user-centric services. This survey paper examines the advances in the area of the analysis of tie strength in social networks. We study the dimensions of tie strength and review the key predictive features for each dimension. We, then, undertake a comparative study of methodologies to model tie strength and examine the key findings. Finally, we discuss open issues and challenges in specifying tie strength.

Despite significant progress in the development of neural-symbolic frameworks, the question of how to integrate a neural and a symbolic system in a compositional manner remains open. Our work seeks to fill this gap by treating these two systems as black boxes to be integrated as modules into a single architecture, without making assumptions on their internal structure and semantics. Instead, we expect only that each module exposes certain methods for accessing the functions that the module implements: the symbolic module exposes a deduction method for computing the function's output on a given input, and an abduction method for computing the function's inputs for a given output; the neural module exposes a deduction method for computing the function's output on a given input, and an induction method for updating the function given input-output training instances. We are, then, able to show that a symbolic module --- with any choice for syntax and semantics, as long as the deduction and abduction methods are exposed --- can be cleanly integrated with a neural module, and facilitate the latter's efficient training, achieving empirical performance that exceeds that of previous work.

Past work on the task of identifying the geographic focus of news-stories has established that state-of-the-art performance can be achieved by using existing crowdsourced knowledge-bases. In this work we demonstrate that a further refinement of those knowledge-bases through an additional round of crowdsourcing can lead to improved performance on the aforementioned task. Our proposed methodology views existing knowledge-bases as collections of arguments in support of particular inferences in terms of the geographic focus of a given news-story. The refinement that we propose is to associate these arguments with weights - computed through crowdsourcing - in terms of how strongly they support their inference. The empirical results that we present establish the superior performance of this approach compared to the one using the original knowledge-base.

Abstract In the past decade, the Winograd schema challenge (WSC) has become a central aspect of the research community as a novel litmus test. Consequently, the WSC has spurred research interest because it can be seen as the means to understand human behavior. In this regard, the development of new techniques has made possible the usage of Winograd schemas in various fields, such as the design of novel forms of CAPTCHAs. Work from the literature that established a baseline for human adult performance on the WSC has shown that not all schemas are the same, meaning that they could potentially be categorized according to their perceived hardness for humans. In this regard, this hardness metric could be used in future challenges or in the WSC CAPTCHA service to differentiate between Winograd schemas. Recent work of ours has shown that this could be achieved via the design of an automated system that is able to output the hardness indexes of Winograd schemas, albeit with limitations regarding the number of schemas it could be applied on. This paper adds to previous research by presenting a new system that is based on machine learning, able to output the hardness of any Winograd schema faster and more accurately than any other previously used method. Our developed system, which works within two different approaches, namely the random forest and deep learning (LSTM-based), is ready to be used as an extension of any other system that aims to differentiate between Winograd schemas, according to their perceived hardness for humans. At the same time, along with our developed system we extend previous work by presenting the results of a large-scale experiment that shows how human performance varies across Winograd schemas.

The Winograd Schema Challenge—the task of resolving pronouns in certain carefully-constructed sentences —has recently been proposed as a basis for a novel form of CAPTCHAs. Such uses of the task necessitate the availability of a large, and presumably continuously-replenished, collection of available Winograd Schemas, which goes beyond what human experts can reasonably develop by themselves. Towards tackling this issue, we introduce Winventor, the first, to our knowledge, system that attempts to fully automate the development of Winograd Schemas, or at least to considerably help humans in this development task. Beyond describing the system, the paper presents a series of three studies that demonstrate, respectively, Winventor’s ability to replicate existingWinograd Schemas from the literature, automatically develop reasonableWinograd Schemas from scratch, and aid humans in developing Winograd Schemas by post-processing the system’s suggestions.

Abstract We present Web-STAR, an online platform for story understanding built on top of the STAR reasoning engine for STory comprehension through ARgumentation. The platform includes a web-based integrated development environment, integration with the STAR system, and a web service infrastructure to support integration with other systems that rely on story understanding functionality to complete their tasks. The platform also delivers a number of “social” features, including a community repository for public story sharing with a built-in commenting system, and tools for collaborative story editing that can be used for team development projects and for educational purposes.

We consider the problem of identifying the geographic focus of a document. Unlike some previous work on this problem, we do not expect the document to explicitly mention the target region, making our problem one of inference or prediction, rather than one of identification. Further, we seek to tackle the problem without appealing to specialized geographic information resources like gazetteers or atlases, but employ general-purpose knowledge bases and ontologies like ConceptNet and YAGO. We propose certain natural strategies towards addressing the problem, and show that the GeoMantis system that implements these strategies outperforms an existing state-of-the-art system, when compared on documents whose target region (country, in particular) is not explicitly mentioned or is obscured. Our results give evidence that using general-purpose knowledge bases and ontologies can, in certain cases, outperform even specialized tools.

This paper presents a novel application of argumentation for automated Story Comprehension (SC). It uses argumentation to develop a computational approach for SC as this is understood and studied in psychology. Argumentation provides uniform solutions to various representational and reasoning problems required for SC such as the frame, ramification, and qualification problems, as well as the problem of contrapositive reasoning with default information. The grounded semantics of argumentation provides a suitable basis for the construction and revision of comprehension models, through the synthesis of the explicit information from the narrative in the text with the implicit (in the reader's mind) common sense world knowledge pertaining to the topic(s) of the story given in the text. We report on the empirical evaluation of the approach through a prototype system and its ability to capture both the majority and the variability of understanding of stories by human readers. This application of argumentation can provide an important test-bed for the more general development of computational argumentation.

Collaboration in nature is often illustrated through the collective behavior of ants. Although entomological studies have shown that certain goals are well-served by this collective behavior, the extent of what such a collaboration can achieve is not immediately clear. We extend past work that has argued that ants are, in principle, able to collectively compute logical circuits, and illustrate through a simulation that indeed such computations can be carried out meaningfully and robustly in situations that involve complex interactions between the ants.

The ability to predict, or at least recognize, the state of the world that an action brings about, is a central feature of autonomous agents. We propose, herein, a formal framework within which we investigate whether this ability can be autonomously learned. The framework makes explicit certain premises that we contend are central in such a learning task: (i) slow sensors may prevent the sensing of an action's direct effects during learning; (ii) predictions need to be made reliably in future and novel situations. We initiate in this work a thorough investigation of the conditions under which learning is or is not feasible. Despite the very strong negative learnability results that we obtain, we also identify interesting special cases where learning is feasible and useful.

A biologically and physically plausible model for ants and pheromones is proposed. It is argued that the mechanisms described in this model are sufficiently powerful to reproduce the necessary components of universal computation. The claim is supported by illustrating the feasibility of designing arbitrary logic circuits, showing that the interactions of ants and pheromones lead to the expected behavior, and presenting computer simulation results to verify the circuits' working. The conclusions of this study can be taken as evidence that coherent deterministic and centralized computation can emerge from the collective behavior of simple distributed Markovian processes such as those followed by biological ants, but also, more generally, by artificial agents with limited computational and communication abilities.

We propose a framework that brings together two major forms of default reasoning in Artificial Intelligence: applying default property classification rules in static domains, and default persistence of properties in temporal domains. Particular attention is paid to the central problem of qualification. We illustrate how previous semantics developed independently for the two separate forms of default reasoning naturally lead to the integration that we propose, and how this gives rise to domains where different types of knowledge interact and qualify each other while preserving elaboration tolerance.

Bloom filter based algorithms have proven successful as very efficient technique to reduce communication costs of database joins in a distributed setting. However, the full potential of bloom filters has not yet been exploited. Especially in the case of multi-joins, where the data is distributed among several sites, additional optimization opportunities arise, which require new bloom filter operations and computations. In this paper, we present these extensions and point out how they improve the performance of such distributed joins. While the paper focuses on efficient join computation, the described extensions are applicable to a wide range of usages, where bloom filters are facilitated for compressed set representation.

We present the first design for a fully expressive iterative combinatorial exchange (ICE). The exchange incorporates a tree-based bidding language that is concise and expressive for CEs. Bidders specify lower and upper bounds on their value for different trades. These bounds allow price discovery and useful preference elicitation in early rounds, and allow termination with an efficient trade despite partial information on bidder valuations. All computation in the exchange is carefully optimized to exploit the structure of the bid-trees and to avoid enumerating trades. A proxied interpretation of a revealed-preference activity rule ensures progress across rounds. A VCG-based payment scheme that has been shown to mitigate opportunities for bargaining and strategic behavior is used to determine final payments. The exchange is fully implemented and in a validation phase.