Artificial Intelligence Agents Research Articles

Reflective Artificial Intelligence

As artificial intelligence (AI) technology advances, we increasingly delegate mental tasks to machines. However, today’s AI systems usually do these tasks with an unusual imbalance of insight and understanding: new, deeper insights are present, yet many important qualities that a human mind would have previously brought to the activity are utterly absent. Therefore, it is crucial to ask which features of minds have we replicated, which are missing, and if that matters. One core feature that humans bring to tasks, when dealing with the ambiguity, emergent knowledge, and social context presented by the world, is reflection. Yet this capability is completely missing from current mainstream AI. In this paper we ask what reflective AI might look like. Then, drawing on notions of reflection in complex systems, cognitive science, and agents, we sketch an architecture for reflective AI agents, and highlight ways forward.

Psychological and Brain Responses to Artificial Intelligence's Violation of Community Ethics.

Human moral reactions to artificial intelligence (AI) agents' behavior constitute an important aspect of modern-day human-AI relationships. Although previous studies have mainly focused on autonomy ethics, this study investigates how individuals judge AI agents' violations of community ethics (including betrayals and subversions) compared with human violations. Participants' behavioral responses, event-related potentials (ERPs), and individual differences were assessed. Behavioral findings reveal that participants rated AI agents' community-violating actions less morally negative than human transgressions, possibly because AI agents are commonly perceived as having less agency than human adults. The ERP N1 component showed the same pattern with moral rating scores, indicating the modulation effect of human-AI differences on initial moral intuitions. Moreover, the level of social withdrawal correlated with a smaller N1 in the human condition but not in the AI condition. The N2 and P2 components were sensitive to the difference between the loyalty/betrayal and authority/subversion domains but not human/AI differences. Individual levels of moral sense and autistic traits also influenced behavioral data, especially on the loyalty/betrayal domain. In our opinion, these findings offer insights for predicting moral responses to AI agents and guiding ethical AI development aligned with human moral values.

How does AI agent (vs. IVR system) service failure impact customer purchase behavior: mediating effect of customer involvement

ABSTRACT In recent years, the use of artificial intelligence (AI) agents in customer service has become a prominent area of study, focusing on their influence over consumer behavior. However, there is a noticeable gap in the current literature regarding AI agent service failures during product information consultations and their effects on customer purchase behavior. The present study, grounded in attribution theory, investigates the effects of various types of self-service failures in both Interactive Voice Response (IVR) systems and AI agents on customer involvement and subsequent purchase behavior. The findings indicate that customers experiencing service failures with AI agents during the consultation stage are more inclined to make purchase behavior after human-led recovery compared to those encountering IVR system failures. Additionally, customer involvement is found to play a pivotal role as a mediator between types of service failures and purchase behavior. The results of this work may offer valuable theoretical and practical insights.

MICIL: Multiple-Instance Class-Incremental Learning for skin cancer whole slide images

Artificial intelligence (AI) agents encounter the problem of catastrophic forgetting when they are trained in sequentially with new data batches. This issue poses a barrier to the implementation of AI-based models in tasks that involve ongoing evolution, such as cancer prediction. Moreover, whole slide images (WSI) play a crucial role in cancer management, and their automated analysis has become increasingly popular in assisting pathologists during the diagnosis process. Incremental learning (IL) techniques aim to develop algorithms capable of retaining previously acquired information while also acquiring new insights to predict future data. Deep IL techniques need to address the challenges posed by the gigapixel scale of WSIs, which often necessitates the use of multiple instance learning (MIL) frameworks. In this paper, we introduce an IL algorithm tailored for analyzing WSIs within a MIL paradigm. The proposed Multiple Instance Class-Incremental Learning (MICIL) algorithm combines MIL with class-IL for the first time, allowing for the incremental prediction of multiple skin cancer subtypes from WSIs within a class-IL scenario. Our framework incorporates knowledge distillation and data rehearsal, along with a novel embedding-level distillation, aiming to preserve the latent space at the aggregated WSI level. Results demonstrate the algorithm’s effectiveness in addressing the challenge of balancing IL-specific metrics, such as intransigence and forgetting, and solving the plasticity-stability dilemma.

DeepCraft: imitation learning method in a cointelligent design to production process to deliver architectural scenarios

The recent advancements in digital technologies and artificial intelligence in the architecture, engineering, construction, and operation sector (AECO) have induced high demands on the digital skills of human experts, builders, and workers. At the same time, to satisfy the standards of the production-efficient AECO sector by reducing costs, energy, health risk, material resources, and labor demand through efficient production and construction methods such as design for manufacture and assembly (DfMA), it is necessary to resolve efficiency-related problems in mutual human‒machine collaborations. In this article, a method utilizing artificial intelligence (AI), namely, generative adversarial imitation learning (GAIL), is presented then evaluated in two independent experiments related to the processes of DfMA as an efficient human‒machine collaboration. These experiments include a) training the digital twin of a robot to execute a robotic toolpath according to human gestures and b) the generation of a spatial configuration driven by a human's design intent provided in a demonstration. The framework encompasses human intelligence and creativity, which the AI agent in the learning process observes, understands, learns, and imitates. For both experimental cases, the human demonstration, the agent's training, the toolpath execution, and the assembly configuration process are conducted digitally. Following the scenario generated by an AI agent in a digital space, physical assembly is undertaken by human builders as the next step. The implemented workflow successfully delivers the learned toolpath and scalable spatial assemblies, articulating human intelligence, intuition, and creativity in the cocreative design.

Preserving paradata for accountability of semi-autonomous AI agents in dynamic environments: An archival perspective

This paper proposes the category of real-time artificial intelligence (AI) systems as applications of computerized control systems in dynamic, time-constrained contexts normally managed by human intelligence. Noting the accountability challenges which these systems introduce, the paper posits the need for robust documentation and records capacities within these systems. The paper surveys four real-time AI systems with significant records needs: autonomous vehicles, online content targeting systems, mixed-reality tools for surgical contexts, and digital twin systems in airport facilities management. The paper identifies paradata, or the data leading up to an output in a system's operation, as a key data category necessitating preservation for full transparency in the records generated by these systems. Paradata is defined as “information about the procedure(s) and tools used to create and process information resources, along with information about the persons carrying out those procedures.” Paradata uncovers opaque technological processes underlying the production of other datasets and at a granular level must be identified and preserved to delineate the boundaries between human and system agency in semi-autonomous systems. With a basis in control theory, the paper finally offers a framework for assessing the functions of real-time AI systems' operations and their documentation and records needs.

The Effect of AI Agent Gender on Trust and Grounding

Artificial intelligence (AI) agents are widely used in the retail and distribution industry. The primary objective was to investigate whether the gender of AI agents influences trust and grounding. This paper examined the influence of AI agent gender and brand concepts on trust and grounding within virtual brand spaces. For this purpose, it used two independent variables: brand concept (functional vs. experiential) and AI agent gender (male vs. female). The dependent variables included AI agent trust and grounding. The study revealed that in virtual brand spaces centered around a functional concept, male AI agents generated higher levels of trust than female AI agents, whereas, when focused on an experiential concept, female AI agents induced higher levels of grounding than male AI agents. Furthermore, the findings indicate that the association between customers’ identification with AI agents and recommendations for actual brand purchases is mediated by trust and grounding. These findings support the idea that users who strongly identify with AI agents are more inclined to recommend brand products. By presenting alternatives that foster the establishment and sustenance of a meaningful, sustainable relationship between humans and AI, this study contributes to research on human–computer interactions.

Assisting the infection preventionist: Use of artificial intelligence for health care–associated infection surveillance

Health care-associated infection (HAI) surveillance is vital for safety in health care settings. It helps identify infection risk factors, enhancing patient safety and quality improvement. However, HAI surveillance is complex, demanding specialized knowledge and resources. This study investigates the use of artificial intelligence (AI), particularly generative large language models, to improve HAI surveillance. We assessed 2 AI agents, OpenAI's chatGPT plus (GPT-4) and a Mixtral 8×7b-based local model, for their ability to identify Central Line-Associated Bloodstream Infection (CLABSI) and Catheter-Associated Urinary Tract Infection (CAUTI) from 6 National Health Care Safety Network training scenarios. The complexity of these scenarios was analyzed, and responses were matched against expert opinions. Both AI models accurately identified CLABSI and CAUTI in all scenarios when given clear prompts. Challenges appeared with ambiguous prompts including Arabic numeral dates, abbreviations, and special characters, causing occasional inaccuracies in repeated tests. The study demonstrates AI's potential in accurately identifying HAIs like CLABSI and CAUTI. Clear, specific prompts are crucial for reliable AI responses, highlighting the need for human oversight in AI-assisted HAI surveillance. AI shows promise in enhancing HAI surveillance, potentially streamlining tasks, and freeing health care staff for patient-focused activities. Effective AI use requires user education and ongoing AI model refinement.

On the Motivations to Seek Information From Artificial Intelligence Agents Versus Humans: A Risk Information Seeking and Processing Perspective

This study investigates how anticipating an artificial intelligence agent versus human information source moderates the risk information seeking and processing model. It focuses on a behavioral proxy of seeking intention—how long a participant waited for an online consultant whose identity was manipulated. In two samples ( N1 = 182 students and N2 = 800 mturkers), the source identity consistently moderated the model in two ways: First, informational subjective norms encouraged seeking from humans but discouraged seeking from AI agents. Second, information insufficiency drove favoritism toward humans–when perceived information-gathering capacity was high. When the capacity was low, AI agents were favored.

How artificial intelligence cooperating with agent‐based modeling for urban studies: A systematic review

AbstractAs urbanization accelerates, cities become more complex, coming along with more complex urban issues. Agent‐based model (ABM) is a traditional method to simulate activities in a complex system, which has been widely applied in urban studies. However, due to its rigid initial settings, ABM has been criticized for its lack of intelligence, especially in dealing with modern urban issues. With the success of artificial intelligence (AI) and complexity science, it is generally agreed that ABM can be enhanced with AI agents, a promising technology that can bridge the gaps. For that, this article provides a systematic review, in which 10 subsections correspond to 10 different ways that AI can work with ABM in the methodological framework. The sections include that (1) ABM is Al; (2) ABM provides training data for Al; (3) Al provides data for ABM; (4) ABM is a submodule in the ensemble Al; (5) Al leads an optimization framework with ABM participation; (6) Al tunes ABM initialization parameters; (7) Al provides the environment for ABM; (8) Al aids in choosing the agent's attributes; (9) Al provides behaviors for agents in ABM; (10) Al helps to evaluate the performance of ABM. For each case, some typical works are examined for illustration. Finally, we discuss some of the current limitations and prospects for future development.

A stimulus exposure of 50 ms elicits the uncanny valley effect

The uncanny valley (UV) effect captures the observation that artificial entities with near-human appearances tend to create feelings of eeriness. Researchers have proposed many hypotheses to explain the UV effect, but the visual processing mechanisms of the UV have yet to be fully understood. In the present study, we examined if the UV effect is as accessible in brief stimulus exposures compared to long stimulus exposures (Experiment 1). Forty-one participants, aged 21–31, rated each human-robot face presented for either a brief (50 ms) or long duration (3 s) in terms of attractiveness, eeriness, and humanness (UV indices) in a 7-point Likert scale. We found that brief and long exposures to stimuli generated a similar UV effect. This suggests that the UV effect is accessible at early visual processing. We then examined the effect of exposure duration on the categorisation of visual stimuli in Experiment 2. Thirty-three participants, aged 21–31, categorised faces as either human or robot in a two-alternative forced choice task. Their response accuracy and variance were recorded. We found that brief stimulus exposures generated significantly higher response variation and errors than the long exposure condition. This indicated that participants were more uncertain in categorising faces in the brief exposure condition due to insufficient time. Further comparisons between Experiment 1 and 2 revealed that the eeriest faces were not the hardest to categorise. Overall, these findings indicate (1) that both the UV effect and categorical uncertainty can be elicited through brief stimulus exposure, but (2) that categorical uncertainty is unlikely to cause the UV effect. These findings provide insights towards the perception of robotic faces and implications for the design of robots, androids, avatars, and artificial intelligence agents.

A framework of diversity, equity, and inclusion safeguards for chatbots

Chatbots such as ChatGPT are conversational agents that play an important role in many sectors. They are highly sophisticated artificial intelligence (AI) agents increasingly integrated into organizational functions ranging from customer support to recruitment and professional development. Alongside the increasing adoption of chatbots is an increasing focus on diversity, equity, and inclusion (DEI). This article presents a series of chatbot examples and discusses their associated DEI implications, such as those related to AI bias. From the cases reviewed in this article, we extract a framework of DEI safeguards, which includes input safeguards, design safeguards, and functional safeguards. This framework can be used by those involved in developing AI based chatbots or managing their use to ensure that chatbots support, rather than weaken, organizational DEI initiatives and strategies.

Empathic voice assistants: Enhancing consumer responses in voice commerce

Artificial intelligence (AI)-enabled voice assistants (VAs) are transforming firm-customer interactions but often come across as lacking empathy. This challenge may cause business managers to question the overall effectiveness of VAs in shopping contexts. Recognizing empathy as a core design element in the next generation of VAs and the limits of scenario-based studies in voice commerce, this article investigates how empathy exhibited by an existing AI agent (Alexa) may alter consumer shopping responses. AI empathy moderates the original structural model bridging functional, relational, and social-emotional dimensions. Findings of an individual-session online experiment show higher intentions to delegate tasks, seek decision assistance, and trust recommendations from AI agents perceived as empathic. In contrast to individual shoppers, families respond better to functional VA attributes such as ease of use when AI empathy is present. The results contribute to the literature on AI empathy and conversational commerce while informing managerial AI design decisions.

Blockchain: The Economic and Financial Institution for Autonomous AI?

This paper examines how the combination of artificial intelligence (AI) and blockchain technology can enable autonomous AI agents to engage and execute economic and financial transactions. We critically examine the constraints on AI agents in achieving predefined objectives independently, especially due to their limited access to economic and financial institutions. We argue that AI’s access to these institutions is vital in enhancing its capabilities to augment human productivity. Drawing on the theory of institutional economics, we propose that blockchain provides a solution for creating digital economic and financial institutions, permitting AI to engage with these institutions through the management of private keys. This extends AI’s capabilities to form and execute contracts, participate in marketplaces, and utilize financial services autonomously. The paper encourages further research on AI as a general-purpose technology and blockchain as an institutional technology that can unlock the full capabilities of autonomous AI agents.

ALGORITHMIC LITERACY: Generative Artificial Intelligence Technologies for Data Librarians

INTRODUCTION: Artificial intelligence (AI) is a novel type of library technology. AI technologies and the needs of data librarians are hybrid and symbiotic, because academic libraries must insert AI technologies into their information and data services. Library services need AI to interpret the context of big data.OBJECTIVES: In this context, we explore the use of the the OpenAI Codex, a deep learning model trained on Python code from repositories, to generate code scripts for data librarians. This investigation examines the practices, models, and methodologies for obtaining code script insights from complex code environments linked to AI GPT technologies. METHODS: The proposed AI-powered method aims to assist data librarians in creating code scripts using Python libraries and plugins such as the integrated development environment PyCharm, with additional support from the Machinet AI and Bito AI plugins. The process involves collaboration between the data librarian and the AI agent, with the librarian providing a natural language description of the programming problem and the OpenAI Codex generating the solution code in Python.RESULTS: Five specific web-scraping problems are presented. The scripts demonstrate how to extract data, calculate metrics, and write the results to files.CONCLUSION: Overall, this study highlights the application of AI in assisting data librarians with code script creation for web scraping tasks. AI may be a valuable resource for data librarians dealing with big data challenges on the Web. The possibility of creating Python code with AI is of great value, as AI technologies can help data librarians work with various types of data sources. The Python code in Data Science web scraping projects uses a machine-learning model that can generate human-like code to help create and improve the library service for extracting data from a web collection. The ability of nonprogramming data librarians to use AI technologies facilitates their interactions with all types and data sources. The Python programming language has artificial intelligence modules, packages, and plugins such as the OpenAI Codex, which serialises automation and navigation in web browsers to simulate human behaviour on pages by entering passwords, selecting captcha options, collecting data, and creating different collections of datasets to be viewed.

The Rise of Robotic Process Automation in the Banking Sector: Streamlining Operations and Improving Efficiency

Robot Process Automation (RPA) is a type of business process automation that relies on software robots (bots) or artificial intelligence (AI) agents. This phenomenon is sometimes denoted as software robotics, which should not be conflated with robot software. This study investigates the increasing prevalence of RPA across several sectors, with a specific focus on its use in back-office functions. RPA software, exemplified by platforms like Blue Prism, Automation Anywhere, and UiPath, replicates human-computer interactions in order to automate operations that are repetitive and governed by predefined rules. This technology offers many advantages, including cost reduction, mistake minimization, and risk elimination. This research investigates many domains in which RPA may be used, including credible business transformation, content migrations, web crawling/OSINT, and IT department enablement. Additionally, it emphasizes the significant responsibilities within RPA operations, including process architects, technologists, and personnel involved in continuous support and maintenance. The study includes case studies conducted within the banking industry, which demonstrate the potential of RPA in augmenting both customer happiness and productivity. The market report anticipates substantial expansion in the market for RPA software, whereby industry leaders such as UiPath, Automation Anywhere, and Blue Prism are expected to play a dominant role.

Intelligent wireless resource management in industrial camera systems: Reinforcement Learning-based AI-extension for efficient network utilization

In this paper, we present an innovative method for replacing wired communication within an operational agile production cell with wireless communication, using a minimal number of access points. Our approach utilizes Reinforcement Learning (RL) to optimize the layout and minimize the required number of access points. We introduce an Artificial Intelligence (AI) agent that facilitates cooperative interaction between cameras and access points, optimizing camera stream utilization and radio coverage. This solution benefits from the recent standardization of The fifth generation of mobile communications (5G) Network Resource Management (NRM) features of the User Equipment (UE) attached to the cameras. To accommodate the dynamic nature of the production cell, our optimization algorithms run during cell operation, necessitating extended training times.Our contributions extend beyond the state-of-the-art by offering a Digital Twin (DT) environment equipped with physics and radio propagation capabilities, fast raytracing for radio propagation modeling, curriculum-learning-based reinforcement learning, and a Physical Twin (PT) of the radio-connected camera with self-adjusting scene setup capabilities. This comprehensive proposal combines the efficiency of 5G radio utilization with precise modeling and high-resolution simulations, while also significantly improving training speed.In the Agile Robotics for Industrial Automation Competition (ARIAC) 2020 environment, our methods showed compelling results, with curriculum learning achieving a remarkable 3x speed-up compared to non-curriculum approaches. We also demonstrated that high Quality of Service (QoS) demands can be reduced to just 2% of total time while maintaining consistent productivity Key Performance Indicators (KPIs). Furthermore, we explored the integration of these methods into Internet of Things (IoT) devices with limited performance capabilities. In conclusion, our approach offers significant benefits and savings for IoT devices, networks, edge clouds, and industry players.

Functional structure of reflexive-active systems of artificial heterogeneous intelligent agents

The paper is aimed at developing distributed artificial intelligence systems for relevant simulation of collective solving of practical problems by modeling reflection. The paper presents the basic design principles and typical functional structure of reflexive-active systems of artificial heterogeneous intelligent agents. Agents of such systems mutually model each other’s reflexive positions; dynamically develop tactics and strategies for their behavior that are relevant to the models of their counterparties; if it is necessary during problem solving, attract new agents from the pool of available agents built by various developers and exclude existing ones, rebuilding “on the fly” the composition of agents and the structure of connections between them.

Human divergent exploration capacity for material design: A comparison with artificial intelligence

Applications of artificial intelligence (AI) to material design have attracted increasing attention in recent years. Although AI-aided material design holds great promise for some applications, whether it has surpassed human creativity remains uncertain. The aim of the current study was to compare the divergent exploration capacity of AI with that of humans on a material design task. Human participants were asked to find a high-performance lubricant molecule under searching conditions comparable to a state-of-the-art AI system. Results indicated that, on average, AI was able to find significantly better lubricant molecules. However, the best molecule found by AI fell short of the best molecule found by a human participant. Furthermore, the structural characteristics of the molecules found by AI and human participants differed significantly. These findings suggest that a state-of-the-art AI system is capable of surpassing human divergent exploration capacity in material design, as in other fields in which AI has advanced. Nevertheless, our results also demonstrate that human intelligence and AI can play complementary roles in covering a broader search space. This investigation opens up new possibilities for collaborative systems involving both AI agents and humans in material design.

Principles of ontophylogenetic development of artificial general intelligence systems based on multi-agent neurocognitive architectures

The purpose of the study is to study the possibilities of multigenerational optimization of behavior control systems for agents of general artificial intelligence capable of independently solving a universal range of tasks in a real environment. The main principles of ontophylogenetic synthesis of control systems for agents of general artificial intelligence based on multi-agent neurocognitive architectures have been developed. Methods and algorithms for synthesizing the phenotypes of control systems of intelligent agents according to their genotypes are proposed. A software package for simulating the processes of ontophylogenetic synthesis of multi-agent neurocognitive architectures has been developed and experiments have been carried out to create phenotypes of intelligent agents based on them. A complex genome of an intelligent agent has been developed, the features of a multichromosome genetic algorithm for organizing calculations in the paradigm of multigenerational optimization of multiagent neurocognitive architectures have been established and substantiated. It is shown that multigenerational optimization of the multi-agent neurocognitive architecture of intelligent agents can contribute to the achievement of adaptive resistance to the operating conditions of a general artificial intelligence agent, provide the synthesis of its suboptimal structural and functional scheme, accelerate learning and algorithms for finding solutions to a universal range of problems solved by this agent in its ecological niche.