AIC 2022, Abstracts of Accepted Papers

D1-1-1: Toward An Artificial Cognitive System To Assist Caregivers In Decision-Making For Persons Living With Dementia

Wellington Pacheco Ferreira and Walter Teixeira Junior.
Abstract: Dementia is a term used to describe a collection of neurodegenerative disorders that impact people across the world. The symptoms of the disease progression include recent memory loss and impairment of cognitive functions that create challenges to individuals living with dementia performing daily tasks, making them dependent on caregivers. Due to social and economic matters, caregivers can be either professional or informal, and in some cases, they are members of the patient's family. Investigations suggest that caring for individuals living with neurodegenerative disease is stressful physically, mentally, and emotionally. This study aims to suggest a model of an artificial cognitive system to assist caregivers in dementia care. The methodology applied was design science research (DSR). We have no knowledge of any work that attempted to implement a similar approach, which corroborates with the belief that this is still a cutting-edge technology and novel interdisciplinary area. Our proof-of-concept tried involving artificial cognitive aspects to assistance. It handled the data of the patient to provide information from reliable sources to non-professional caregiver's decision-making regardless of the disease stage. However, it is in a nascent state of development and requires further research to be applied in real scenarios. [PDF]

D1-1-2: Understanding eye-tracking in virtual reality

Maurice Lamb, Estela Pérez Luque and Erik Billing.                                                                                    
[PDF: Extended Abstract]

D1-1-3:Making sense with social robots: Extending the landscape of investigation in HRI

Valentina Fantasia, Ingar Brinck and Christian Balkenius
Abstract: The aim of this position paper is to propose a reflection on how to account for and investigate the many ways in which interaction between robots and humans requires co-ordination, negotiation or reformulation of meanings emerging in the ongoing interaction. Towards this aim, we argue, a perspective shift may be needed: to frame interactions as social engagements whose meaning can only be understood by the standpoint of those participating in it. We first present research on psychological benchmarks and design patterns for sociality in the HRI field. Then we provide arguments for including a new interactional element currently missing in the literature: participatory sense-making processes. As we will argue, such elements can be conceived and operationalised both as a relational benchmark as well as an interactional pattern, therefore proving useful for HRI research. [PDF]

D1-1-4:Appreciation of Symbolic Attributes in Machine Perception

Mohamadreza Faridghasemnia
Abstract: In this position paper, we want to attract attention to the importance of symbolic attributes in machine perception. We discuss the benefits of a perception system that not only recognizes the category of objects, but also recognizes many other aspects of objects. [PDF]

D1-2-1:Quantifying the Impact of Predicate Similarities on Knowledge Graph Triple Embeddings

Alexander Kalinowski and Yuan An
Abstract: In devising methods for representing knowledge graph triples in low-dimensional spaces, care must be taken to quantify the similarities between all components, especially the predicate components common to all triples. Unfortunately, knowledge graph benchmarks do not come equipped with scores indicating the semantic similarity between two arbitrary triples. To proxy these scores, we introduce a weakly supervised method we call PTSS, or pairwise triple similarity scoring. A neural model then utilizes this information to update triple representations. We conduct experiments using this method by substituting three methods for predicate relatedness measures: linear algebraic similarities of predicate embeddings, predicate frequency/inverse predicate frequency, and KL-divergence of predicate distributions. We analyze the information captured by these approaches and their impacts when utilized as weak supervision signals for triple representations to test the hypothesis that these approaches reflect a notion of cognitive similarity. Our findings indicate a combined model using scores driven from neural embeddings for entities and fact distributions for predicates achieves the best results, highlighting the efficacy of combining neural and distributional approaches and suggests this pattern of combination may be fruitful in other cognitively inspired AI solutions. [PDF]

D1-2-2: Generative Logic Models for Data-Based Symbolic Reasoning            

Hiroyuki Kido
Abstract: Acquiring knowledge from data and reasoning with the obtained knowledge are both essential processes of successful logical systems. However, most current logical systems assume different algorithms for the two processes. The separation causes serious problems such as knowledge acquisition bottleneck, grounding and commonsense reasoning. This paper gives a simple probabilistic model unifying the two processes. It formalises how data generate models of formal logic and the models generate the truth values of logical formulae. The generated models and truth values are shown to be consistent with maximum likelihood estimation and Fenstad's theorem, respectively. Probabilistic reasoning on logical formulae is shown to be a reasonable alternative to a logical consequence relation and a paraconsistent consequence relation. This paper contributes to data-based reasoning with linear complexity. [PDF]

D1-2-3:Embodied Affordance Grounding using Semantic Simulations and Neural-Symbolic Reasoning: An Overview of the PlayGround Project

Andreas Persson and Amy Loutfi
Abstract: In this paper, we present a synopsis of the PlayGround project. Through neural-symbolic learning and reasoning, the PlayGround project assumes that high-level concepts and reasoning processes can be used to advance both symbol grounding and object affordance inference. However, a prerequisite for reasoning about objects and their affordances is integrated object representations that concurrently maintain symbolic values (e.g., high-level concepts), and sub-symbolic features (e.g., spatial aspects of objects). Integrated representations that, preferably, should be based upon neural-symbolic computation such that neural-symbolic models can, subsequently, be used for high-level reasoning processes. Nevertheless, reasoning processes for symbol grounding and affordance inference often require multiple inference steps. Taking inspiration from the cognitive prospects in simulation semantics, the PlayGround project further presumes that these reasoning processes can be simulated by neural rendering complementary to high-level reasoning processes. [PDF]

D1-2-4:Assessing the Time Efficiency of Ethical Algorithms

Jakob Stenseke and Christian Balkenius
Abstract: Artificial moral agents must not only be able to make competent ethical decisions, but they must do so effectively. This paper explores how ethical theory and algorithmic design impact computational efficiency by assessing the time cost of ethical algorithms. We create a model of an ethical environment and conduct experiments on three different ethical algorithms in order to compare computational benefits and disadvantages of deontology and consequentialism respectively. The experimental results highlight the close relationship between ethical theory, algorithmic design, and resource costs, and our work provides an important starting-point for the further examination of these relations. Lastly, we introduce the concept of moral tractability as a venue for future work. [PDF]

D2-1-1:Experimental design and facets of evidence for computational theory of mind

Joel Michelson, Deepayan Sanyal, James Ainooson, Yuan Yang and Maithilee Kunda
Abstract: The competitive feeding paradigm is one of several experimental setups intended to test whether non-verbal subjects possess skills related to Theory of Mind. Competitive feeding focuses on the relationship between seeing and knowing. In this paper, we describe a highly-customizeable implementation of the competitive feeding paradigm for computational agents in a gridworld environment. We explore various modifications to the setup including shared rewards, alternate sequences of timed events, and asymmetrical values, that allow us to replicate a wide breadth of tests designed to study the social cognition skills of humans and animals. Finally, we describe how this paradigm can be expanded upon and used as a benchmark test to investigate social reasoning in artificially intelligent models [PDF]

D2-1-2:The Source of Desire: Personal Identity as a Drive for Agent Behavior

Ursula Addison
Abstract: The central component in the design of our artificial agent behavior generation system, la VIDA, is its drive mechanism. This drive, which we have named an identity profile, is a collection of roles, values, beliefs, and attitudes analogous to and inspired by the human identity. This identity profile acts as an intrinsic source of motivation to drive long-term, autonomous agent behavior. In this paper we explore the representation and function of an early design for the identity profile. We also discuss how the identity profile interacts with a personality inventory and commonsense knowledge graph to assign traits and associated strengths to the agent. These values can then be mapped to three behavior intensities to augment agent actions. [PDF]

D2-1-3: Cognitive Mimetics - SMT Model                    

Antero Karvonen and Pertti Saariluoma  
Abstract: Cognitively inspired intelligent systems have recently been attracting renewed attention. However, understanding this approach in terms of design processes seems to be still in the early stages. We propose that something essential in the endeavor can understood through the lens of mimetic design, and in particular cognitive mimetics. We will frame the discussion through a model of mimetic design consisting of a source, a target and a mapping relation. At this abstract level, it provides a canvas into which different types of cognitively inspired design processes can be mapped - descriptively and prescriptively depending on purposes. It provides a starting point against which theoretical, ontological, and methodological questions, commitments, and specifications can be made explicitly. [PDF]

D2-1-4: On Proactive Human-AI systems                                                            

Jasmin Grosinger
Abstract: With a growing number of AI systems and robots sharing the environment of humans, the need to define and investigate the particular topic of artificial proactivity is greater than ever. This position paper advocates the importance of this endeavor and starts the work by giving an initial definition of proactivity for artificial agents, analyzing the cognitive abilities necessary to create proactive agent behavior and suggests a categorization of approaches in different types of proactivity. [PDF]

D2-1-5:Road map for cognitively-inspired artificial systems

Sara Mahmoud and Alice Plebe
Abstract: Nature has been a great source of inspiration for many inventions and theories. One of the major benefits for this inspiration is perceiving the impossible as possible. The inception of the AI field was no exception with cognitively-inspired approaches with a dream of having an intelligent system that thinks as a human. However, this journey of human intelligence into machine intelligence has been rough and more challenging that resulted in the separation of AI from cognitive studies. In this article, we highlight the main challenges and opportunities for cognitive inspiration for AI development. We then break down the source of inspiration into four abstraction levels in which the researcher may place an inspiration from. These levels then contribute into three main stages for modeling the AI system. The two dimensional mapping from cognitive levels into modeling stages and the relation between them aims to assist the process of cognitively-inspired approaches. [PDF]

D2-2-1:A Conceptual Chronicle of Solving Raven's Progressive Matrices Computationally

Yuan Yang, Deepayan Sanyal, Joel Michelson, James Ainooson and Maithilee Kunda
Abstract: Matrix reasoning or geometric analogy problems, like those found on the widely used Raven's Progressive Matrices test of intelligence, have been used as a challenge for machine intelligence since the early work of Evans in the 1960s. While AI research on the RPM has gone through dramatic shifts alongside other AI advances, including a dramatic rise of machine-learning-based approaches in the last five years, many of these studies have progressed in relatively siloed research lines, making it difficult to compare different approaches and judge progress in the field as a whole. This paper intends to provide a framework for understanding the different lines of work in this research field. In particular, we reviewed 50+ computational models for solving RPM or RPM-like problems and collated them into a linear conceptual framework to help researchers navigate across these diverse research paradigms. We also provide instructions on other resources such as problem/data sets and necessary background knowledge of RPM. [PDF]

D3-A1-1:Commonsense Spatial Reasoning for Visually Intelligent Agents

Agnese Chiatti, Gianluca Bardaro, Enrico Motta and Enrico Daga
Abstract: Service robots are expected to reliably make sense of complex, fast-changing environments. From a cognitive standpoint, they need the appropriate reasoning capabilities and background knowledge required to exhibit human-like Visual Intelligence. In particular, our prior work has shown that i) commonsense reasoning is a necessary capability for Visual Intelligence and also that ii) commonsense reasoning crucially requires the ability to reason about spatial relations between objects in the world. In this paper, we first recap our approach to Visual Intelligence in robotics, which is based on a hybrid architecture integrating a deep learning component with commonsense reasoning. We then present a framework for spatial reasoning, which has been designed to support the commonsense reasoning component in our architecture. Differently from prior approaches to qualitative spatial reasoning in robotics, the proposed framework is robust to variations in the robot's viewpoint and object orientation. In the paper, we also show how this formally-defined framework can be operationalised in an off-the-shelf spatial database. [PDF]

D3-A1-2:Inherently Interpretable Knowledge Representation for a Trustworthy Artificially Intelligent Agent Teaming with Humans in Industrial Environments

Vedran Galetic and Alistair Nottle
Abstract: Embodied artificially intelligent agents teaming with humans in industrial environments must be safe and trustworthy, their behaviour predictable, and their rationale explainable. In addressing these extremely wide requirements, we take the knowledge representation angle. Adopting Gärdenfors's Conceptual Space framework, learnt concepts are represented as regions across inherently interpretable quality dimensions, while classification of instances proceeds using a simple derivative model assuming fuzzy category membership. In our use case from the manufacturing domain, the quality dimensions consist of physical properties retrievable from the agent's sensors and utilisation properties from crowdsourced commonsense knowledge. This heterogeneous property decomposition approach allows for flexible concept acquisition and manipulation, particularly useful for industrial settings often characterised by highly specific artefacts and thus data scarcity, which may impact the effectiveness of the state-of-the-art data-hungry — and typically opaque — computer-vision based approaches. [PDF]

D3-A1-3:Towards Learning Abstractions via Reinforcement Learning

Erik Jergéus, Leo Karlsson Oinonen, Emil Carlsson and Moa Johansson
Abstract: In this paper we take the first steps in studying a new approach to synthesis of efficient communication schemes in multi-agent systems, trained via reinforcement learning. We combine symbolic methods with machine learning, in what is referred to as a neuro-symbolic system. The agents are not restricted to only use initial primitives: reinforcement learning is interleaved with steps to extend the current language with novel higher-level concepts, allowing generalisation and more informative communication via shorter messages. We demonstrate that this approach allow agents to converge more quickly on a small collaborative construction task. [PDF]

D3-A1-4: A situation-calculus based model of linguistic context                    

Richard Scherl
Abstract: This paper develops a logical model based on the situation-calculus of the role of indexicality in creating the linguistic context and how the context changes through the actions of the agent. It looks at indexicals of person, place, and time by showing how the egocentric position is linked towards different objective "maps" of space and time. Multi-participant discourse sets up relations between between the turn units of conversation, while narrative discourse demands relations between the context and embedded context. [PDF]

D3-2-1:Argumentation and Causal Models in Human-Machine Interaction: A Round Trip

Yann Munro, Isabelle Bloch, Mohamed Chetouani, Marie-Jeanne Lesot and Catherine Pelachaud
Abstract: In the field of explainable artificial intelligence (XAI), causal models and abstract argumentation frameworks constitute two formal approaches that provide definitions of the notion of explanation. These symbolic approaches rely on logical formalisms to reason by abduction or to search for causalities, from the formal modeling of a problem or a situation. They are designed to satisfy properties that have been established as necessary based on the study of human-human explanations. As a consequence they appear to be particularly interesting for human-machine interactions as well. In this paper, we show the equivalence between a particular type of causal models, that we call argumentative causal graphs (ACG), and abstract argumentation frameworks. We also propose a transformation between these two systems and look at how one definition of an explanation in the argumentation theory is transposed when moving to ACG. To illustrate our proposition, we use a very simplified version of a screening agent for COVID-19. [PDF]

D3-2-2:Towards Robotic Minds: Dynamic Interpretation and Schemata Recombination

Stevan Tomic
Abstract: Humans excel at interpreting the world dynamically and forming simplifying abstractions. Moreover, depending on the context, humans can re-interpret one situation in terms of another, which allows us to assign different meanings to (same) physical objects. With this ability, we can re-use existing knowledge, understand the behaviors of others, use language with metaphors, and learn and reason in novel situations. These seem to be prevalent in humans in contrast to other animals or, at least, current computational systems. Thus, a cognitive-inspired computational framework would need to consider dynamic interpretation and abstraction as the main design principle. This work presents such a framework focusing on the problem of dynamic interpretation. [PDF]

D3-2-3:Exploring Swedish & English fastText Embeddings

Tosin Adewumi, Foteini Simistira Liwicki and Marcus Liwicki
Abstract: In this paper, we show that embeddings from relatively smaller corpora sometimes outperform those from larger corpora and we introduce a new Swedish analogy test set and make it publicly available. To achieve good performance in Natural Language Processing (NLP) downstream tasks, several factors play important roles: dataset size, the right hyper-parameters, and well-trained embeddings. We utilize the fastText tool for our experiments. We evaluate both the Swedish and English embeddings that we created using intrinsic evaluation (including analogy & Spearman correlation) and compare them with 2 common, publicly available embeddings. Our English continuous Bag-of-Words (CBoW)-negative sampling embedding shows better performance compared to the publicly available GoogleNews version. We also describe the relationship between NLP and cognitive science. We contribute the embeddings for research or other useful purposes by publicly releasing them. [PDF]