Human Decision Research Articles

Explainability increases trust resilience in intelligent agents.

Even though artificial intelligence (AI)-based systems typically outperform human decision-makers, they are not immune to errors, leading users to lose trust in them and be less likely to use them again-a phenomenon known as algorithm aversion. The purpose of the present research was to investigate whether explainable AI (XAI) could function as a viable strategy to counter algorithm aversion. We conducted two experiments to examine how XAI influences users' willingness to continue using AI-based systems when these systems exhibit errors. The results showed that, following the observation of algorithms erring, the inclination of users to delegate decisions to or follow advice from intelligent agents significantly decreased compared to the period before the errors were revealed. However, the explainability effectively mitigated this decline, with users in the XAI condition being more likely to continue utilizing intelligent agents for subsequent tasks after seeing algorithms erring than those in the non-XAI condition. We further found that the explainability could reduce users' decision regret, and the decrease in decision regret mediated the relationship between the explainability and re-use behaviour. These findings underscore the adaptive function of XAI in alleviating negative user experiences and maintaining user trust in the context of imperfect AI.

Ethical Concerns Upon Artificial Intelligence Empowered Human Resource Management: A Qualitative Study among Middle-level Managers from Beijing Technology Companies

The evolution of artificial intelligence (AI) in organizational management has significantly enhanced operational efficiency. However, it has introduced ethical challenges in human resource management. Between January to March 2024, this study conducted 21 semi-structured interviews with middle-level managers from high-tech companies in Beijing. Through word frequency analysis, the study found that key topics among the managers were “company,” “data,” “system,” and “problem,” with “AI” frequently recurring in the discussions. Sentiment analysis revealed generally favorable attitudes toward AI in human resource management (AI-HRM), along with nuanced emotional expressions such as inquiry, introspection, recommendation, challenge, adaptation, resistance, and indifference. The sentiment distribution of keywords aligned with topic trends. Thematic analysis identified key ethical concerns in AI-HRM, including issues related to data collection and utilization, human versus machine decision-making, quantitative versus qualitative assessment methodologies, the balance between fairness and efficiency, the need for trustworthy, explainable, and transparent AI, and the oversight of AI-HRM. This study contributes to the ethical investigation of AI-HRM from the perspective of middle-level managers, highlighting themes that are critical for understanding the theory, application, and future development of AI-HRM.

How mariculture expansion is dewilding the ocean and its inhabitants.

The world's oceans are largely free from intensive farming, but momentum to intensify and expand mariculture-the cultivation of aquatic organisms in the ocean-is growing. Despite optimism that mariculture will create economic and nutritional benefits for humans, it can also generate a host of risks, including environmental degradation, harms to wildlife integrity and welfare, captivity effects, and shifts in how humans view the nonhuman world. Collectively, we refer to these four types of risks as "dewilding." In this systematic review, we searched Scopus and Web of Science for recent literature documenting mariculture's dewilding impacts to organize and collate this evidence under one unified framework. We find that mariculture's dewilding impacts are consistently documented, though often in isolation, and that captivity and conceptual dewilding impacts are recognized as potential harms far less than impacts on the environment and wildlife. Future work examining mariculture's dewilding impacts will be paramount to guiding human decision-making and activity going forward.

Intelligent Borders: Exploring the Suitability of Artificial Intelligence Systems in Refugee Status Determination Under International Law

Abstract This article assesses the potential use of Artificial Intelligence (AI) in Refugee Status Determination, exploring how AI systems, including biometrics, predictive analytics, and emotion recognition, could support or replace human decision-making in determining refugee status. While AI could speed up and improve the objectivity of Refugee Status Determination procedures, the article raises concerns about its limitations in assessing the subjective fear of persecution, a critical element of refugee claims. The inability of AI systems to fully account for emotional and personal nuances leads to the possibility of discriminatory practices, particularly when AI is used without human oversight. The study highlights how AI could exacerbate power imbalances between asylum-seekers and States, transforming these technologies into tools for anti-immigration policies. It therefore calls for caution in the use of AI in Refugee Status Determination procedures and stresses that human interaction remains essential for the fair assessment of refugee claims. This article concludes that while AI can assist in certain aspects of the Refugee Status Determination procedures, its full implementation without careful legal and human rights considerations poses significant risks.

Boosting: Empowering Citizens with Behavioral Science.

Behavioral public policy came to the fore with the introduction of nudging, which aims to steer behavior while maintaining freedom of choice. Responding to critiques of nudging (e.g., that it does not promote agency and relies on benevolent choice architects), other behavioral policy approaches focus on empowering citizens. Here we review boosting, a behavioral policy approach that aims to foster people's agency, self-control, and ability to make informed decisions. It is grounded in evidence from behavioral science showing that human decision making is not as notoriously flawed as the nudging approach assumes. We argue that addressing the challenges of our time-such as climate change, pandemics, and the threats to liberal democracies and human autonomy posed by digital technologies and choice architectures-calls for fostering capable and engaged citizens as a first line of response to complement slower, systemic approaches.

Predicting and Understanding Human Action Decisions: Insights from Large Language Models and Cognitive Instance-Based Learning

Large Language Models (LLMs) excel in tasks from translation to complex reasoning. For AI systems to help effectively, understanding and predicting human behavior and biases is essential. However, it remains an open question whether LLMs can achieve this goal. This paper addresses this gap by leveraging the reasoning and generative capabilities of LLMs to predict human behavior in two sequential decision-making tasks. These tasks involve balancing between exploratory and exploitative actions and handling delayed feedback, which is essential for simulating real-life decision processes. We compare the performance of LLMs with a cognitive instance-based learning (IBL) model, which imitates human experiential decision-making. Our findings indicate that LLMs excel at rapidly incorporating feedback to enhance prediction accuracy. In contrast, the IBL model better accounts for human exploratory behaviors and effectively captures loss aversion bias — the tendency to choose a sub-optimal goal with fewer step-cost penalties rather than exploring to find the optimal choice, even with limited experience. The results highlight the benefits of integrating LLMs with cognitive architectures, suggesting that this synergy could enhance the modeling and understanding of complex human decision-making patterns.

GeoGail: A Model-based Imitation Learning Framework for Human Trajectory Synthesizing

Synthesized human trajectories are crucial for a large number of applications. Existing solutions are mainly based on the generative adversarial network (GAN), which is limited due to the lack of modeling the human decision-making process. In this paper, we propose a novel imitation learning based method to synthesize human trajectories. This model utilizes a novel semantics-based interaction mechanism between the decision-making strategy and visitations to diverse geographical locations to model them in the semantic domain in a uniform manner. To augment the modeling ability to the real-world human decision-making policy, we propose a feature extraction model to extract the internal latent factors of variation of different individuals, and then propose a novel self-attention based policy net to capture the long-term correlation of mobility and decision-making patterns. Then, to better reward users’ mobility behavior, we propose a novel multi-scale reward net combined with mutual information to model the instant reward, long-term reward, and individual characteristics in a cohesive manner. Extensive experimental results on two real-world trajectory datasets show that our proposed model can synthesize the most high-quality trajectory data compared with six state-of-the-art baselines in terms of a number of key usability metrics, and can well support practical applications based on trajectory data, demonstrating its effectiveness. Furthermore, our proposed method can learn explainable knowledge automatically from data, including explainable statistical features of trajectories and statistical relation between decision-making policy and features.

Mix and Match: Characterizing Heterogeneous Human Behavior in AI-assisted Decision Making

AI-assisted decision-making systems hold immense potential to enhance human judgment, but their effectiveness is often hindered by a lack of understanding of the diverse ways in which humans take AI recommendations. Current research frequently relies on simplified, ``one-size-fits-all'' models to characterize an average human decision-maker, thus failing to capture the heterogeneity of people's decision-making behavior when incorporating AI assistance. To address this, we propose Mix and Match (M&M), a novel computational framework that explicitly models the diversity of human decision-makers and their unique patterns of relying on AI assistance. M&M represents the population of decision-makers as a mixture of distinct decision-making processes, with each process corresponding to a specific type of decision-maker. This approach enables us to infer latent behavioral patterns from limited data of human decisions under AI assistance, offering valuable insights into the cognitive processes underlying human-AI collaboration. Using real-world behavioral data, our empirical evaluation demonstrates that M&M consistently outperforms baseline methods in predicting human decision behavior. Furthermore, through a detailed analysis of the decision-maker types identified in our framework, we provide quantitative insights into nuanced patterns of how different individuals adopt AI recommendations. These findings offer implications for designing personalized and effective AI systems based on the diverse landscape of human behavior patterns in AI-assisted decision-making across various domains.

The effects of artificial intelligence and victims’ deservingness information on citizens’ blame attribution towards administrative errors

ABSTRACT This study examines how citizens attribute blame to government authorities for administrative errors made by artificial intelligence (AI) compared to human decision-makers. Based on blame attribution theory, we conducted a vignette-based survey experiment with 1,098 Chinese citizens, revealing that respondents assign less blame for errors caused by AI or AI-assisted decisions. Additionally, disclosing victims’ deservingness information heightened blame attribution. These findings contribute to the literature on administrative accountability, highlighting how citizens respond to AI-related errors and informing the growing use of AI in public sector decision-making.

Behavioral Economics Concepts and the Q.E. Method

In Behavioral Economics, the notion of perfect rationality is replaced by a more nuanced understanding that recognizes the limits of human decision-making, such as bounded rationality, willpower, self-interest, and attention. Instead of expecting individuals to always act in perfectly rational ways, models in Behavioral Economics are judged by the accuracy of their predictions rather than the realism of their assumptions. This perspective aligns well with the Quantification of Quality Data (Q.E.) method, which also prioritizes the accuracy and integrity of the model's outcomes over the simplicity of its assumptions. The Q.E. method begins with a solid theoretical framework, where a hypothesis is carefully crafted to define the scope and objectives of the analysis. This initial step is crucial because it sets the foundation for the mathematical determination that follows, ensuring that the analysis is rooted in a clear theoretical context rather than being driven solely by empirical data. By generating values for independent variables within a defined range through randomization, the Q.E. method allows for a comprehensive exploration of how these variables influence the model. This process typically involves the use of multiple mathematical equations to fully capture the complex behavior of the model. A key feature of the Q.E. method is its feedback loop, which plays a critical role in maintaining consistency between the mathematical model and the underlying theoretical principles. Through iterative hypothesis testing, model adjustment, and continuous feedback, the Q.E. method ensures that the mathematical model not only aligns with the theory but also accurately reflects the behavioral nuances that are central to Behavioral Economics. This rigorous approach allows for the refinement of the initial hypothesis and the confirmation of the model’s predictive accuracy, ensuring that the quantification of quality data is both theoretically sound and empirically valid. In this way, the Q.E. method supports the development of Behavioral Economic models that are robust, reliable, and reflective of real-world decision-making processes.

Административные решения алгоритмов

Examples of algorithms being involved in administrative practice are multiplying. However, the process of their use remains unsettled. For the legal evaluation of the results of algorithms’ decisions and subsequent legal regulation, it is important to distinguish between automated decision-making systems and machine learning systems. But if the prospects for the use of algorithms by both courts and the executive branch seem obvious, such use should certainly take into account possible risks and disadvantages. This is especially true for machine learning algorithms. First, human decision-making inevitably leads to the problem of scoring, which has not only an ethical but also a legal dimension. Second, there is the problem of algorithm errors. A statistical decision, even based on a working algorithm, can be wrong. Statistics using probability calculus, even when applied to a mass of cases, leads to probable rather than certain conclusions. This means that the personalized decision of a machine learning algorithm cannot be final, but must be “sanctioned” by a human. Hence the inevitability of human control.

Algorithms advise, humans decide: the evidential role of the patient preference predictor

An AI-based ‘patient preference predictor’ (PPP) is a proposed method for guiding healthcare decisions for patients who lack decision-making capacity. The proposal is to use correlations between sociodemographic data and...

Machine Learning, Deep Learning, Artificial Intelligence and Aesthetic Plastic Surgery: A Qualitative Systematic Review.

This systematic review aims to assess the use of machine learning, deep learning, and artificial intelligence in aesthetic plastic surgery. This qualitative systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses reporting guideline. To analyze quality risk-of-bias assessment of all included articles, we used the ROBINS-I tool for non-randomized studies. We searched for studies with the following MeSH terms: Machine Learning OR Deep Learning OR Artificial intelligence AND Plastic surgery on MEDLINE/PubMed, EMBASE, and Cochrane Library, from inception until July 2024 without any filter applied. A total of 2,148 studies were screened and 41 were fully reviewed. We conducted article extraction, screening, and full text review using the rayyan tool. Eighteen studies were ultimately included in this review, describing the use of machine learning, deep learning and artificial intelligence in aesthetic plastic surgery. All studies were published from 2019 to 2024. Articles varied regarding the population studied, type of machine learning (ML), Deep Learning Model (DLM), Artificial Intelligence (AI) used, and aesthetic plastic surgery type. Of the eighteen studies, we included the following aesthetic plastic surgeries: augmentation mastopexy, breast augmentation, reduction mammoplasty, rhinoplasty, facial rejuvenation surgery, including facelift surgery; blepharoplasty, and body contouring. Image-based with AI, ML, and DLMs algorithms were used in these studies to improve human decision-making and identified factors associated with postoperative complications. AI, ML, and DL algorithms offer immense potential to transform the aesthetic plastic surgery field. By meticulously analyzing patient data, these technologies may, in the future, help optimize treatment plans, predict potential complications, and more clearly elucidate patient concerns, improving their ability to make informed decisions. The drawback, as with preoperative surgical simulation, is that patients may see an AI-generated image that is to their liking, but impossible to achieve; great care is needed when using such tools in order to not create unrealistic expectations. Ultimately, the old plastic surgery adage of ''under-promise and over-deliver'' will continue to hold true, at least for the foreseeable future. This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266 . Study registration A review protocol for this systematic review was registered at PROSPERO CRD42024567461.

Advice from artificial intelligence: a review and practical implications.

Despite considerable behavioral and organizational research on advice from human advisors, and despite the increasing study of artificial intelligence (AI) in organizational research, workplace-related applications, and popular discourse, an interdisciplinary review of advice from AI (vs. human) advisors has yet to be undertaken. We argue that the increasing adoption of AI to augment human decision-making would benefit from a framework that can characterize such interactions. Thus, the current research invokes judgment and decision-making research on advice from human advisors and uses a conceptual "fit"-based model to: (1) summarize how the characteristics of the AI advisor, human decision-maker, and advice environment influence advice exchanges and outcomes (including informed speculation about the durability of such findings in light of rapid advances in AI technology), (2) delineate future research directions (along with specific predictions), and (3) provide practical implications involving the use of AI advice by human decision-makers in applied settings.

How distributed subcortical integration of reward and threat may inform subsequent approach-avoidance decisions.

Healthy and successful living involves carefully navigating rewarding and threatening situations by balancing approach and avoidance behaviours. Excessive avoidance to evade potential threats often leads to forfeiting potential rewards. However, little is known about how reward and threat information is integrated neurally to inform approach or avoidance decisions. In this preregistered study, participants (Nbehaviour=31, 17F; NMRI=28, 15F) made approach-avoidance decisions under varying reward (monetary gains) and threat (electrical stimulations) prospects during functional MRI scanning. In contrast to theorized parallel subcortical processing of reward and threat information before cortical integration, Bayesian Multivariate Multilevel analyses revealed subcortical reward and threat integration prior to indicating approach-avoidance decisions. This integration occurred in the ventral striatum, thalamus, and bed nucleus of the stria terminalis (BNST). When reward was low, risk-diminishing avoidance decisions dominated, which was linked to more positive tracking of threat magnitude prior to indicating avoidance than approach decisions across these regions. In contrast, the amygdala exhibited dual sensitivity to reward and threat. While anticipating outcomes of risky approach decisions, we observed positive tracking of threat magnitude within the salience network (dorsal anterior cingulate cortex, thalamus, periaqueductal gray, BNST). Conversely, after risk-diminishing avoidance, characterized by reduced reward prospects, we observed more negative tracking of reward magnitude in the ventromedial prefrontal cortex and ventral striatum. These findings shed light on the temporal dynamics of approach-avoidance decision-making. Importantly, they demonstrate the role of subcortical integration of reward and threat information in balancing approach and avoidance, challenging theories positing predominantly separate subcortical processing prior to cortical integration.Significance statement When deciding whether to approach or avoid situations, our decision-making involves balancing potential rewards and threats. Widespread theories of decision-making in humans propose parallel processing of reward and threat information in subcortical regions, followed by cortical integration. Challenging these notions, we found evidence for dual and integrated processing of reward and threat in subcortical regions during decision-making. In contrast, after decision-making, we observed the expected parallel processing while anticipating decision outcomes. These findings advance our understanding of approach-avoidance decision-making processes, opposing traditional views that segregate brain regions as predominantly reward-sensitive or threat-sensitive, thereby paving the way for a more nuanced perspective that takes into account the stage of decision-making.

All-In-One Optoelectronic Transistors for Bio-Inspired Visual System.

Visual perception has profound effects on human decision-making and emotional responses. Replicating the functions of the human visual system through device development has been a constant pursuit in recent years. However, to fully simulate the various functions of the human visual system, it is often necessary to integrate multiple devices with different functions, resulting in complex, large-volume device structures and increased power consumption. Here, an optoelectronic transistor with comprehensive visual functions is introduced. By coupling diverse photoreceptive properties of the channel and electrical regulation through charge injection/ferroelectric switching from the hafnium-based gate, the devices can simulate functions of both photoreceptors in the retina and synapses in the visual cortex. A device array is constructed to confirm the perceptual functions of cone and rod cells. Subsequently, color discrimination and recognition for color images are achieved by combining the tunable perception and synapse functions. Then an intelligent traffic judgment system with this all-in-one device is developed, which is capable of making judgments and decisions regarding traffic signals and pedestrian movements. This work provides a potential solution for developing compact and efficient devices for the next-generation bio-inspired visual system.

Real-World Geopolitical Influences on In-Game Financial Behaviour

This paper explores how real-world geopolitics can influence financial decision-making in virtual environments. The study primarily focuses on trading dynamics in a pandemic simulation game. The players manage resources and trade with each other to save their population. We compared two groups: one playing with made-up country names (control) and another with real country names (test). The players were not permitted to communicate outside of the resources the game provides and were finally ranked based on the number of months survived and trade volume. Our findings reveal two distinct player archetypes, one focusing on maximising trade volume and a second focusing on resource conservation. Real country names compared to color-coded labels seem to delay the perception of competition and reduce overall trade volume, suggesting increased immersion and a more conservative approach towards resource management. Observations from gameplay footage reveal instances of potential bias, with players forming diplomatic ties based on national identities. The data collected suggests that real-world geopolitics can influence financial behaviour in virtual environments as hypothesised. It also explores virtual environments to study how political narratives and dynamics affect human decision-making. The findings offer practical insights for game developers regarding fostering player engagement, mitigating bias, and creating fair economies. While parallels with real-world resource trading and allocation during pandemics were noted, further analysis would be required to explore these connections.

A large language model-based agent for wayfinding: simulation of spatial perception and memory

ABSTRACT Using agents to simulate human wayfinding processes helps to understand human decision-making and spatial cognition. The performance of large language models (LLMs) in agent-based wayfinding is still unclear. To address this, this study harnessed the thinking and reasoning capabilities of LLMs by designing an agent wayfinding framework based on cellular automata. This framework provides environmental descriptions in text form to the LLMs, enabling them to perceive the wayfinding environment and navigate within it. We simulated different spatial memory conditions for the LLM, including three types of spatial memories and one condition with no spatial memory, to observe its wayfinding behavior in a real-world setting. Multiple experiments were designed to compare the wayfinding performance of the LLM under different spatial memories, and 32 human participants conducted the same wayfinding experiments for reference. The results indicate that the LLM autonomously explores the environment and displays spatial reasoning abilities similar to humans. Furthermore, the LLM can develop reasonable wayfinding strategies to complete the tasks. This framework allows agents with different spatial memories to accurately reach their destinations. This research provides new ideas for utilizing large language models to construct more intelligent, efficient, and flexible wayfinding agents, which can help advance the application of agent technology in fields such as navigation and robotics.

A Study on the Development of an Integrated Automatic Allocation Algorithms for Multilayer Defense System Intercept Missiles

This study aims to propose optimal missile interception allocation strategies for constructing a multi-layered missile defense system in situations where a large number of ballistic missiles are launched simultaneously. Specifically, we consider the threat level of ballistic missiles, the importance of defense zones, and the asset value of interception missiles, and develop a damage consumption index based on these factors. Through an algorithm that minimizes the sum of damage consumption indices for intercepting missiles defending upper and lower layers, we allocate them in optimal combinations to achieve maximum effectiveness with minimal cost. Additionally, we propose a decision-making system based on algorithmic integration rather than human decision-making, particularly in complex operational scenarios, to ensure systematic decision-making and maximize efficiency.

Emotion-Driven Moral Evaluation: A Mechanistic Study Based on the Drift-Diffusion Model.

Moral evaluation is identified as the first stage in the theory of moral judgment, and academics believe that it may align with the social intuitionist model. This study aims to prove that the model's emotional dominance hypothesis applies to moral evaluation by presenting a computational decision-making model that mathematically formalizes this emotional dominance decision-making process. We also compared different types of valence evaluation tasks to test the emotional priority hypothesis. We used a convenience sampling method to randomly recruit 30 enrolled college students. The drift-diffusion model was employed to analyze reaction times for words with various emotional and moral valences Additionally, we designed different valence evaluation tasks based on the response relevance hypothesis and evaluated the processing order through reaction time comparisons. The analysis revealed that the emotional mechanism of immoral evaluation differs from moral evaluation. An increase in emotional valence accelerates the speed of evidence accumulation (v) for moral evaluation (M = 1.21, 0.2% < 0 < 99.8%) but lowers decision caution (a) in immoral evaluation (M = -0.64, 96.1% < 0 < 3.9%). In contrast, moral valence does not have a significant influence on evaluation processes (v, M = -0.28, 72.1% < 0 < 27.9%; a, M = -0.32, 79.3% < 0 < 20.7%). Furthermore, We found no significant difference in reaction times between moral and immoral words in the emotional evaluation task (F(1,29) = 0.55, p = 0.464, partial η2 = 0.02), but a significant difference existed in the moral evaluation task (F(1,29) = 17.99, p < 0.001, partial η2 = 0.38), indicating that the tendency of relatively fast immoral evaluation in emotional evaluation tasks may be caused by emotional priority. Our findings support the intuitive model's emotional dominance hypothesis and introduce a new emotional mechanism into moral evaluation. This study clarifies the distinct emotional processes in moral and immoral evaluations, fills a gap in the research on moral evaluation, and offers insights into human decision-making in moral contexts.