AI Agents Research Articles

System of Operations on Associative Heterarchical Memory

Text processing in natural language remains an important task for the field of development of artificial intelligence methods and tools. Since the twentieth century, artificial intelligence methods have been divided into two paradigms — top-down and bottom-up. The methods of the ascending paradigm are difficult to interpret in the form of the output of natural language, and the methods of the descending paradigm are difficult to actualize information. Taking into account the authors approach to the construction of artificial intelligence agents, the processing of natural language must be performed on two levels: on the lower level, using methods of the bottom-up paradigm, and on the upper level, using symbolic methods of the top-down paradigm. The authors of the article have already introduced a new mathematical formalism based on the notion of a hypergraph — associative heterarchical memory (AH-memory). Such memory should simplify the process of natural language processing with new technologies. Earlier the authors group has thoroughly analyzed the problem of symbol binding in the application to АН-memory and its structure. In the first paper, abstract symbol binding was performed using multi-serial integration, eventually converting the primary symbols received by the program into integrated abstract symbols. The second paper provided a comprehensive description of the AH-memory in the form of formulas, explanations of them, and their corresponding diagrams. Although there are many possible modules to use, the developer working with AH-memory should choose those parts of AH-memory which are required for successful and efficient functioning of the AI agent. The article will be of interest to developers of artificial intelligence methods and tools, mathematicians and specialists in natural language processing.

Design of a Hyper-Casual Futsal Mobile Game Using a Machine-Learned AI Agent-Player

Mobile games continue to gain popularity, and their revenues are increasing accordingly. However, due to the inherent constraints of small screen sizes and restrictions of computing, it has been considered challenging to simulate the complex gameplay of soccer games. To this end, this paper aims to design and develop a simplified version of a five vs. five hyper-casual futsal game with only three player positions: goalkeeper, striker, and defender. It also tests a demo game to verify whether it is possible to implement an AI agent−player for each position to machine-learn and to run on a mobile device. A demo game with an AI agent−player was simulated using both PPO and SAC algorithms, and the feasibility and stability of the algorithms were compared. The results showed that each AI agent−player achieved the assigned objectives for each position and successfully machine-learned. When the algorithms were compared, the SAC algorithm showed a more stable state than the PPO algorithm when SAC directed the gameplay and interactive AI techniques. This paper shows the great potential of the application of machine-learned AI agent−players for soccer simulators on mobile platforms.

Influence of Anthropomorphic Agent on Human Empathy Through Games

The social acceptance of AI agents, including intelligent virtual agents and physical robots, is becoming more important for the integration of AI into human society. Although the agents used in human society share various tasks with humans, their cooperation may frequently reduce the task performance. One way to improve the relationship between humans and AI agents is to have humans empathize with the agents. By empathizing, humans feel positively and kindly toward agents, which makes it easier to accept them. In this study, we focus on tasks in which humans and agents have various interactions together, and we investigate the properties of agents that significantly influence human empathy toward the agents. To investigate the effects of task content, difficulty, task completion, and an agent’s expression on human empathy, two experiments were conducted. The results of the two experiments showed that human empathy toward the agent was difficult to maintain with only task factors, and that the agent’s expression was able to maintain human empathy. In addition, a higher task difficulty reduced the decrease in human empathy, regardless of task content. These results demonstrate that an AI agent’s properties play an important role in helping humans accept them.

Data Formulator: AI-Powered Concept-Driven Visualization Authoring.

With most modern visualization tools, authors need to transform their data into tidy formats to create visualizations they want. Because this requires experience with programming or separate data processing tools, data transformation remains a barrier in visualization authoring. To address this challenge, we present a new visualization paradigm, concept binding, that separates high-level visualization intents and low-level data transformation steps, leveraging an AI agent. We realize this paradigm in Data Formulator, an interactive visualization authoring tool. With Data Formulator, authors first define data concepts they plan to visualize using natural languages or examples, and then bind them to visual channels. Data Formulator then dispatches its AI-agent to automatically transform the input data to surface these concepts and generate desired visualizations. When presenting the results (transformed table and output visualizations) from the AI agent, Data Formulator provides feedback to help authors inspect and understand them. A user study with 10 participants shows that participants could learn and use Data Formulator to create visualizations that involve challenging data transformations, and presents interesting future research directions.

Economic Expectations and an AI Agent

Economic Expectations and an AI Agent

A shape of play to come: Exploring children's play and imaginaries with robots and AI

We are rapidly moving into an era where AI and robots are part of everyday interactions in society and education, and there are immense discussions today about current and future technologies. Still, children are often not included in this discussion, while there is much to learn from current uses and children's understandings of AI and robotics. The study is based on a seven-month ethnographical work that details the implementation of a robot in two preschool groups of children aged 1–2 and 3–5 (n = 38). The study descriptively combines a framework for children's play analysis with explorative qualitative child interviews (n = 6) with the 3-5-year-olds to examine how children play with the robot and their thinking about a future with robots and AI. The results show how children's play with robots spans all of Hughes's (2011) sixteen play types and integrates robots into play in ways specific to child-robot interaction. The interviews indicate that children have well-formed knowledge about the current uses of robots and AI and elaborate imaginaries about a future with them, including critical boundaries toward robots and AI agents. The evidence shows emerging ways children relate to these. The potential of including children's actions and voices in the ongoing societal and educational debates on AI is discussed.

How and Why the Use of AI (vs. Human) Agents Influences Sales Conversion Rate

How and Why the Use of AI (vs. Human) Agents Influences Sales Conversion Rate

When the (Bayesian) Ideal Is Not Ideal

Bayesian epistemologists support the norms of probabilism and conditionalization using Dutch book and accuracy arguments. These arguments assume that rationality requires agents to maximize practical or epistemic value in every doxastic state, which is evaluated from a subjective point of view (e.g., the agent’s expectancy of value). The accuracy arguments also presuppose that agents are opinionated. The goal of this paper is to discuss the assumptions of these arguments, including the measure of epistemic value. I have designed AI agents based on the Bayesian model and a nonmonotonic framework and tested how they achieve practical and epistemic value in conditions in which an alternative set of assumptions holds. In one of the tested conditions, the nonmonotonic agent, which is not opinionated and fulfills neither probabilism nor conditionalization, outperforms the Bayesian in the measure of epistemic value that I argue for in the paper (α -value). I discuss the consequences of these results for the epistemology of rationality.

I Shop Therefore I Am: The Artificial Consumer

Can Artificial Intelligence replace humans not only as producers of goods and content, but also as consumers of commodities? AI agents have so far been associated mainly with the production of goods and content—that is, supply—which is only one side of economic processes. The other side is consumption, which drives demand. Is an Artificial Consumer (AC) conceivable? Can artificial consumers support a human–artificial hybrid artificial economy? Could such an economy help increase GDP? Such questions require exploration of the connection between consciousness, consumption, value, economics, and AI. This admittedly speculative paper discusses the potential consequences of introducing artificial consumers. What place would be left in the economic system for humans if AI took over both supply and demand? The final frontier for AI could be our economy: I shop therefore I am.

Vero: An accessible method for studying human–AI teamwork

Despite the recognized need to prepare for a future of human–AI collaboration, the technical skills necessary to develop and deploy AI systems are considerable, making such research difficult to perform without specialized knowledge. To make human–AI collaboration research more accessible, we developed a novel experimental method that combines a standard video conferencing platform, a set of animations, and Wizard of Oz methods to simulate a group interaction with an AI teammate. Through a case study, we demonstrate the flexibility and ease of deployment of this approach. We also provide evidence that the method creates a highly believable experience of interacting with an AI agent. By detailing this method, we hope that researchers regardless of background can replicate it to more easily answer questions that will inform the design and development of future human–AI collaboration technologies.

Cocreative Interaction: Somax2 and the REACH Project

Abstract Somax2 is an artificial intelligence (AI)-based multiagent system for human–machine “coimprovisation” that generates stylistically coherent streams while continuously listening and adapting to musicians or other agents. The model on which it is based can be used with little configuration to interact with humans in full autonomy, but it also allows fine real-time control of its generative processes and interaction strategies, closer in this case to a “smart” digital instrument. An offspring of the Omax system, conceived at the Institut de Recherche et Coordination Acoustique/Musique (IRCAM), the Somax2 environment is part of the European Research Council Raising Cocreativity in Cyber–Human Musicianship (REACH) project, which studies distributed creativity as a general template for symbiotic interaction between humans and digital systems. It fosters mixed musical reality involving cocreative AI agents. The REACH project puts forward the idea that cocreativity in cyber–human systems results from the emergence of complex joint behavior, produced by interaction and featuring cross-learning mechanisms. Somax2 is a first step toward this ideal, and already shows life-size achievements. This article describes Somax2 extensively, from its theoretical model to its system architecture, through its listening and learning strategies, representation spaces, and interaction policies.

Artificial intelligence insights into hippocampal processing.

Advances in artificial intelligence, machine learning, and deep neural networks have led to new discoveries in human and animal learning and intelligence. A recent artificial intelligence agent in the DeepMind family, muZero, can complete a variety of tasks with limited information about the world in which it is operating and with high uncertainty about features of current and future space. To perform, muZero uses only three functions that are general yet specific enough to allow learning across a variety of tasks without overgeneralization across different contexts. Similarly, humans and animals are able to learn and improve in complex environments while transferring learning from other contexts and without overgeneralizing. In particular, the mammalian extrahippocampal system (eHPCS) can guide spatial decision making while simultaneously encoding and processing spatial and contextual information. Like muZero, the eHPCS is also able to adjust contextual representations depending on the degree and significance of environmental changes and environmental cues. In this opinion, we will argue that the muZero functions parallel those of the hippocampal system. We will show that the different components of the muZero model provide a framework for thinking about generalizable learning in the eHPCS, and that the evaluation of how transitions in cell representations occur between similar and distinct contexts can be informed by advances in artificial intelligence agents such as muZero. We additionally explain how advances in AI agents will provide frameworks and predictions by which to investigate the expected link between state changes and neuronal firing. Specifically, we will discuss testable predictions about the eHPCS, including the functions of replay and remapping, informed by the mechanisms behind muZero learning. We conclude with additional ways in which agents such as muZero can aid in illuminating prospective questions about neural functioning, as well as how these agents may shed light on potential expected answers.

Teachable Conversational Agents for Crowdwork: Effects on Performance and Trust

Traditional crowdsourcing has mostly been viewed as requester-worker interaction where requesters publish tasks to solicit input from human crowdworkers. While most of this research area is catered towards the interest of requesters, we view this workflow as a teacher-learner interaction scenario where one or more human-teachers solve Human Intelligence Tasks to train machine learners. In this work, we explore how teachable machine learners can impact their human-teachers, and whether they form a trustable relation that can be relied upon for task delegation in the context of crowdsourcing. Specifically, we focus our work on teachable agents that learn to classify news articles while also guiding the teaching process through conversational interventions. In a two-part study, where several crowd workers individually teach the agent, we investigate whether this learning by teaching approach benefits human-machine collaboration, and whether it leads to trustworthy AI agents that crowd workers would delegate tasks to. Results demonstrate the benefits of the learning by teaching approach, in terms of perceived usefulness for crowdworkers, and the dynamics of trust built through the teacher-learner interaction.

Development of Chatbot-Psychologist: Dataset, Architecture, Design and Chatbot in Use

Virtual assistants are software applications that allow users to have conversations with the software in the same way as they would in real life. Creating functional conversational agents have been the most challenging aspect of working with AI ever since the field was first introduced. Virtual conversational agents' primary responsibility is to properly interpret natural communication and reply in an acceptable manner, in spite of the fact that they may do a range of activities. In the past, AI agents were created either via the use of handwritten rules and commands or by straightforward statistical approaches. End-to-end Automation has mostly supplanted these models because of its superior potential for learning new skills. The encoder-decoder deep learning approach is currently the method that has the greatest amount of interest for simulating conversations. The domain of linguistic understanding served as a source of motivation for the development of this concept. In this article, we present an overview o the findings from our study into the development of an immersive digital conversational agent that might potentially provide sufferers psychological aid. We used Rasa NLU approach, which is based on NLP practices, in order to construct and train the chatbot. The findings of the study indicated that providing appropriate replies to patients' questions and concerns during conversations had a prediction accuracy about 75 percent.

Toward a 'Standard Model' of Machine Learning

Machine learning (ML) is about computational methods that enable machines to learn concepts from experience. In handling a wide variety of experience ranging from data instances, knowledge, and constraints to rewards, adversaries, and lifelong interaction in an ever-growing spectrum of tasks, contemporary ML/AI (artificial intelligence) research has resulted in a multitude of learning paradigms and methodologies. Despite the continual progresses on all different fronts, the disparate, narrowly focused methods also make standardized, composable, and reusable development of ML approaches difficult, and preclude the opportunity to build AI agents that panoramically learn from all types of experience. This article presents a standardized ML formalism, in particular a 'standard equation' of the learning objective, that offers a unifying understanding of many important ML algorithms in the supervised, unsupervised, knowledge-constrained, reinforcement, adversarial, and online learning paradigms, respectively—those diverse algorithms are encompassed as special cases due to different choices of modeling components. The framework also provides guidance for mechanical design of new ML approaches and serves as a promising vehicle toward panoramic machine learning with all experience.

"I Want To See How Smart This AI Really Is": Player Mental Model Development of an Adversarial AI Player

Understanding players' mental models are crucial for game designers who wish to successfully integrate player-AI interactions into their game. However, game designers face the difficult challenge of anticipating how players model these AI agents during gameplay and how they may change their mental models with experience. In this work, we conduct a qualitative study to examine how a pair of players develop mental models of an adversarial AI player during gameplay in the multiplayer drawing game iNNk. We conducted ten gameplay sessions in which two players (n = 20, 10 pairs) worked together to defeat an AI player. As a result of our analysis, we uncovered two dominant dimensions that describe players' mental model development (i.e., focus and style). The first dimension describes the focus of development which refers to what players pay attention to for the development of their mental model (i.e., top-down vs. bottom-up focus). The second dimension describes the differences in the style of development, which refers to how players integrate new information into their mental model (i.e., systematic vs. reactive style). In our preliminary framework, we further note how players process a change when a discrepancy occurs, which we observed occur through comparisons (i.e., compare to other systems, compare to gameplay, compare to self). We offer these results as a preliminary framework for player mental model development to help game designers anticipate how different players may model adversarial AI players during gameplay.

Bringing Together Ergonomic Concepts and Cognitive Mechanisms for Human—AI Agents Cooperation

The deployment of artificial intelligence from experimental settings to concrete applications implies to consider the social aspects of the environment and consequently to conceive the interaction between humans and computers endowed with the aim of being partners in action. This article proposes a review of the research initiatives regarding human-artificial agents interaction, including eXplainable Artificial Intelligence (XAI) and HRI/HCI. We argue that even if vocabulary and approaches are different, the concepts converge on the necessity for the artificial agents to provide an accurate mental model of their behavior to the humans they are interacting with. This has different implications depending on whether we consider a tool/user interaction or a cooperation interaction—which is far less documented despite being at the heart of the future concepts of autonomous vehicles. From this observation, the article uses the cognitive science corpus on joint-action to raise finer cognitive mechanisms proved to be essential for human joint-action which could be considered as cognitive requirements for future artificial agents, including shared task representation and mentalization. Finally, interactions content hypotheses are arisen to satisfy the identified mechanisms, including the ability for the artificial agent to elicit its intentions and to trigger mentalization toward them from the human cooperators.

Conversational AI for multi-agent communication in Natural Language

Research at the Interaction Lab focuses on human-agent communication using conversational Natural Language. The ultimate goal is to create systems where humans and AI agents (including embodied robots) can spontaneously form teams and coordinate shared tasks through the use of Natural Language conversation as a universal communication interface. This paper first introduces machine learning approaches to problems in conversational AI in general, where computational agents must coordinate with humans to solve tasks using conversational Natural Language. It also covers some of the practical systems developed in the Interaction Lab, ranging from speech interfaces on smart speakers to embodied robots interacting using visually grounded language. In several cases communication between multiple agents is addressed. The paper surveys the central research problems addressed here, the approaches developed, and our main results. Some key open research questions and directions are then discussed, leading towards a future vision of conversational, collaborative multi-agent systems.

Tourists and AI: A political ideology perspective

The tourism industry has adopted AI agents as substitutes for human contact. We examine how tourists respond to AI in hotel service settings during a pandemic. Four studies show that services featuring human interaction is preferred to AI enabled interaction. Moreover, subjective happiness is identified as the underlying causal mechanism that drives this effect. We provide further nuanced insights by showing that politically conservative tourists discriminate between service agent types more. Theoretically we contribute to a more fine-grained understanding of tourists' responses towards AI by examining an affective and an individual difference factor (political ideology). Practically, marketers are encouraged to incorporate tourists' ideologies and psychographics into segmentation, targeting, and positioning considerations, thereby enhancing their marketing effectiveness.

Adapt and overcome: Perceptions of adaptive autonomous agents for human-AI teaming

Rapid advances in AI technologies have caused teams to explore the use of AI agents as full, active members of the team. The complex environments that teams occupy require human team members to constantly adapt their behaviors, and thus the ability of AI teammates to similarly adapt to changing situations significantly enhances the team’s chances to succeed. In order to design such agents, it is important that we understand not only how to identify the amount of autonomous control AI agents have over their decisions, but also how changes to this control cognitively affects the rest of the team. Professional organizations often break their work cycles into phases that set limits on the team members’ actions, and we propose that a similar process could be used to define the autonomy levels of AI teammates. Cyber incident response is an ideal context for this proposal, as we were able to use incident response phases to explore how a team’s work cycle could guide an AI agent’s changing level of autonomy. Using a mixed methods approach, we recruited 103 participants to complete a factorial survey containing ten contextual vignettes focused on an AI teammate’s level of autonomy in incident response contexts, and from these participants we conducted twenty-two follow-on qualitative interviews that further explored how the participants felt an AI agent’s adaptive capabilities would affect team performance and cohesiveness. Our results showed that work cycles can be used to assign autonomy levels to adaptive AI agents based upon the degree of formal processes and predictability of the team’s tasks during the cycle, and that dynamic, human-like adaptation methods are vital to effective human-AI teams. This research provides significant contributions to the HCI community by proposing design recommendations for the development of adaptive autonomous teammates that both enhance Human-AI teams’ productivity and promote positive team dynamics.