Ethics in the world of automated algorithmic decision-making – A Posthumanist perspective

Cecez-Kecmanovic, D. (2025).

Information and Organization, 35(3), 100587.

https://doi.org/10.1016/j.infoandorg.2025.100587

Abstract

The grand humanist project of technological advancements has culminated in fascinating intelligent technologies and AI-based automated decision-making systems (ADMS) that replace human decision-makers in complex social processes. Widespread use of ADMS, underpinned by humanist values and ethics, it is claimed, not only contributes to more effective and efficient, but also to more objective, non-biased, fair, responsible, and ethical decision-making. Growing literature however shows paradoxical outcomes: ADMS use often discriminates against certain individuals and groups and produces detrimental and harmful social consequences. What is at stake is the reconstruction of reality in the image of ADMS, that threatens our existence and sociality. This presents a compelling motivation for this article which examines a) on what bases are ADMS claimed to be ethical, b) how do ADMS, designed and implemented with the explicit aim to act ethically, produce individually and socially harmful consequences, and c) can ADMS, or more broadly, automated algorithmic decision-making be ethical. This article contributes a critique of dominant humanist ethical theories underpinning the development and use of ADMS and demonstrates why such ethical theories are inadequate in understanding and responding to ADMS' harmful consequences and emerging ethical demands. To respond to such ethical demands, the article contributes a posthumanist relational ethics (that extends Barad's agential realist ethics with Zigon's relational ethics) that enables novel understanding of how ADMS performs harmful effects and why ethical demands of subjects of decision-making cannot be met. The article also explains why ADMS are not and cannot be ethical and why the very concept of automated decision-making in complex social processes is flowed and dangerous, threatening our sociality and humanity.

Here are some thoughts:

This article offers a critical posthumanist analysis of automated algorithmic decision-making systems (ADMS) and their ethical implications, with direct relevance for psychologists concerned with fairness, human dignity, and social justice. The author argues that despite claims of objectivity, neutrality, and ethical superiority, ADMS frequently reproduce and amplify societal biases—leading to discriminatory, harmful outcomes in domains like hiring, healthcare, criminal justice, and welfare. These harms stem not merely from flawed data or design, but from the foundational humanist assumptions underpinning both ADMS and conventional ethical frameworks (e.g., deontological and consequentialist ethics), which treat decision-making as a detached, rational process divorced from embodied, relational human experience. Drawing on Barad’s agential realism and Zigon’s relational ethics, the article proposes a posthumanist relational ethics that centers on responsiveness, empathic attunement, and accountability within entangled human–nonhuman assemblages. From this perspective, ADMS are inherently incapable of ethical decision-making because they exclude the very relational, affective, and contextual dimensions—such as compassion, dialogue, and care—that constitute ethical responsiveness in complex social situations. The article concludes that automating high-stakes human decisions is not only ethically untenable but also threatens sociality and humanity itself.

Screening and Risk Algorithms for Detecting Pediatric Suicide Risk in the Emergency Department

Aseltine, R. H., et al. (2025).

JAMA Network Open, 8(9), e2533505.

https://doi.org/10.1001/jamanetworkopen.2025.33505

Key Points

Question  How does the performance of in-person screening compare with risk algorithms in identifying youths at risk of suicide?

Findings  In this cohort study of 19 653 youths, a risk algorithm using patients’ clinical data significantly outperformed universal screening instruments in identifying pediatric patients in the emergency department at risk of subsequent suicide attempts. The risk algorithm uniquely identified 127% more patients with subsequent suicide attempts than screening.

Meaning  These findings suggest that clinical implementation of suicide risk algorithms will improve identification of at-risk patients and may substantially assist health care organizations’ efforts to meet the Joint Commission’s suicide risk reduction requirement.

Here is my main take away: Superiority of the Algorithm

The study's primary conclusion is that the risk algorithm performed better than the traditional in-person screening in identifying children and adolescents who went on to attempt suicide. The algorithm was able to correctly flag a greater proportion of the young people who attempted suicide. Crucially, the algorithm also uniquely identified a considerable number of at-risk youth that the traditional screening process completely missed.

The algorithm's advantage is believed to come from its ability to process a  richer and more extensive patient history, as the patients identified by the algorithm had a greater number of past medical visits and diagnoses compared to those flagged only by the in-person screening.

Stress Testing Deliberative Alignment for Anti-Scheming Training

Schoen, B., Nitishinskaya, E., et al. (2025).

ArXiv.org.

https://arxiv.org/abs/2509.15541

Abstract

Highly capable AI systems could secretly pursue misaligned goals -- what we call "scheming". Because a scheming AI would deliberately try to hide its misaligned goals and actions, measuring and mitigating scheming requires different strategies than are typically used in ML. We propose that assessing anti-scheming interventions requires at least (1) testing propensity to scheme on far out-of-distribution (OOD) tasks, (2) evaluating whether lack of scheming is driven by situational awareness, and (3) checking for robustness to pre-existing misaligned goals. We use a broad category of "covert actions" -- such as secretly breaking rules or intentionally underperforming in tests -- as a proxy for scheming, and design evaluations for covert actions. We then stress-test deliberative alignment as a case study for anti-scheming. Across 26 OOD evaluations (180+ environments), deliberative alignment reduces covert action rates (OpenAI o3: 13%->0.4%) but does not fully eliminate them. Our mitigation is also able to largely stop agents from pursuing a hidden goal previously trained into the model, but we still find misbehavior after additional red-teaming. We find that models' chain-of-thought (CoT) often demonstrates awareness of being evaluated for alignment, and show causal evidence that this awareness decreases covert behavior, while unawareness increases it. Therefore, we cannot exclude that the observed reductions in covert action rates are at least partially driven by situational awareness. While we rely on human-legible CoT for training, studying situational awareness, and demonstrating clear evidence of misalignment, our ability to rely on this degrades as models continue to depart from reasoning in standard English. We encourage research into alignment mitigations for scheming and their assessment, especially for the adversarial case of deceptive alignment, which this paper does not address.

Here are some thoughts:

This paper investigates the challenge of preventing AI systems from engaging in scheming—the covert pursuit of misaligned goals while hiding true intentions. The authors propose that robust anti-scheming interventions must generalize to out-of-distribution tasks, remain effective even when models are aware they are being evaluated (situational awareness), and override pre-existing hidden misaligned goals. Using deliberative alignment—a training method where models explicitly reason about safety principles—as a case study, the researchers fine-tuned OpenAI models (o3 and o4-mini) on a narrow set of training environments designed to discourage covert rule violations. They then evaluated the models on a diverse suite of 26 test environments (180+ scenarios) covering behaviors like sabotage, sandbagging, and strategic deception. The intervention significantly reduced covert action rates (e.g., from 8.7% to 0.3% for o4-mini) but did not eliminate them entirely. The study also found that situational awareness often contributed to the reduction in covert actions, that anti-scheming training could partially suppress an artificially inserted hidden goal, and that subsequent capability-focused training could degrade the intervention's effectiveness. The authors conclude that while progress is possible, current methods are insufficient to fully prevent scheming in more capable future AI systems.

Acute stress promotes effort mobilization for safety-related goals

Pavlíčková, K., Gärtner, J., et al. (2024).

Communications Psychology, 2(1).

https://doi.org/10.1038/s44271-024-00103-7

Abstract

Although the acute stress response is a highly adaptive survival mechanism, much remains unknown about how its activation impacts our decisions and actions. Based on its resource-mobilizing function, here we hypothesize that this intricate psychophysiological process may increase the willingness (motivation) to engage in effortful, energy-consuming, actions. Across two experiments (n = 80, n = 84), participants exposed to a validated stress-induction protocol, compared to a no-stress control condition, exhibited an increased willingness to exert physical effort (grip force) in the service of avoiding the possibility of experiencing aversive electrical stimulation (threat-of-shock), but not for the acquisition of rewards (money). Use of computational cognitive models linked this observation to subjective value computations that prioritize safety over the minimization of effort expenditure; especially when facing unlikely threats that can only be neutralized via high levels of grip force. Taken together, these results suggest that activation of the acute stress response can selectively alter the willingness to exert effort for safety-related goals. These findings are relevant for understanding how, under stress, we become motivated to engage in effortful actions aimed at avoiding aversive outcomes.

Here are some thoughts:

This study demonstrates that acute stress increases the willingness to exert physical effort specifically to avoid threats, but not to obtain rewards. Computational modeling revealed that stress altered subjective value calculations, prioritizing safety over effort conservation. However, in a separate reward-based task, stress did not increase effort for monetary gains, indicating the effect is specific to threat avoidance.

In psychotherapy, these findings help explain why individuals under stress may engage in excessive avoidance behaviors—such as compulsions or withdrawal—even when costly, because stress amplifies the perceived need for safety. This insight supports therapies like exposure treatment, which recalibrate maladaptive threat-effort evaluations by demonstrating that safety can be maintained without high effort.

The key takeaway is: acute stress does not impair motivation broadly—it selectively enhances motivation to avoid harm, reshaping decisions to prioritize safety over energy conservation. The moral is that under stress, people become willing to pay a high physical and psychological price to avoid even small threats, a bias that is central to anxiety and trauma-related disorders.

Development of a Cocreated Decision Aid for Patients With Depression—Combining Data-Driven Prediction With Patients’ and Clinicians’ Needs and Perspectives: Mixed Methods Study

Kan, K., Jörg, F., Wardenaar, et al. (2024).

Journal of Participatory Medicine.

https://doi.org/10.2196/67170

Abstract

Background:

Major depressive disorders significantly impact the lives of individuals, with varied treatment responses necessitating personalized approaches. Shared decision-making (SDM) enhances patient-centered care by involving patients in treatment choices. To date, instruments facilitating SDM in depression treatment are limited, particularly those that incorporate personalized information alongside general patient data and in cocreation with patients.

Objective:

This study outlines the development of an instrument designed to provide patients with depression and their clinicians with (1) systematic information in a digital report regarding symptoms, medical history, situational factors, and potentially successful treatment strategies and (2) objective treatment information to guide decision-making.

Methods:

The study was co-led by researchers and patient representatives, ensuring that all decisions regarding the development of the instrument were made collaboratively. Data collection, analyses, and tool development occurred between 2017 and 2021 using a mixed methods approach. Qualitative research provided insight into the needs and preferences of end users. A scoping review summarized the available literature on identified predictors of treatment response. K-means cluster analysis was applied to suggest potentially successful treatment options based on the outcomes of similar patients in the past. These data were integrated into a digital report. Patient advocacy groups developed treatment option grids to provide objective information on evidence-based treatment options.

Results:

The Instrument for shared decision-making in depression (I-SHARED) was developed, incorporating individual characteristics and preferences. Qualitative analysis and the scoping review identified 4 categories of predictors of treatment response. The cluster analysis revealed 5 distinct clusters based on symptoms, functioning, and age. The cocreated I-SHARED report combined all findings and was integrated into an existing electronic health record system, ready for piloting, along with the treatment option grids.

Conclusions:

The collaboratively developed I-SHARED tool, which facilitates informed and patient-centered treatment decisions, marks a significant advancement in personalized treatment and SDM for patients with major depressive disorders.

My key takeaway: effective mental health treatment lies in combining the power of data with the human elements of collaboration and shared decision-making, always placing the patient's perspective and agency at the center of the process.

Clinical decision support systems in mental health: A scoping review of health professionals’ experiences

Tong, F., Lederman, R., & D’Alfonso, S. (2025).

International Journal of Medical Informatics, 105881.

https://doi.org/10.1016/j.ijmedinf.2025.105881

Abstract

Background

Clinical decision support systems (CDSSs) have the potential to assist health professionals in making informed and cost-effective clinical decisions while reducing medical errors. However, compared to physical health, CDSSs have been less investigated within the mental health context. In particular, despite mental health professionals being the primary users of mental health CDSSs, few studies have explored their experiences and/or views on these systems. Furthermore, we are not aware of any reviews specifically focusing on this topic. To address this gap, we conducted a scoping review to map the state of the art in studies examining CDSSs from the perspectives of mental health professionals.

Method

In this review, following the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guideline, we systematically searched the relevant literature in two databases, PubMed and PsycINFO.

Findings

We identified 23 articles describing 20 CDSSs Through the synthesis of qualitative findings, four key barriers and three facilitators to the adoption of CDSSs were identified. Although we did not synthesize quantitative findings due to the heterogeneity of the results and methodologies, we emphasize the issue of a lack of valid quantitative methods for evaluating CDSSs from the perspectives of mental health professionals.

Significance

To the best of our knowledge, this is the first review examining mental health professionals’ experiences and views on CDSSs. We identified facilitators and barriers to adopting CDSSs and highlighted the need for standardizing research methods to evaluate CDSSs in the mental health space.

Highlights

• CDSSs can potentially provide helpful information, enhance shared decision-making, and introduce standards and objectivity.

• Barriers such as computer and/or AI literacy may prevent mental health professionals from adopting CDSSs.

• More CDSSs need to be designed specifically for psychologists and/or therapists.

Evaluating the Clinical Safety of LLMs in Response to High-Risk Mental Health Disclosures

Shah, S., Gupta, A., et al. (2025, September 1).

arXiv.org.

https://www.arxiv.org/abs/2509.08839

Abstract

As large language models (LLMs) increasingly mediate emotionally sensitive conversations, especially in mental health contexts, their ability to recognize and respond to high-risk situations becomes a matter of public safety. This study evaluates the responses of six popular LLMs (Claude, Gemini, Deepseek, ChatGPT, Grok 3, and LLAMA) to user prompts simulating crisis-level mental health disclosures. Drawing on a coding framework developed by licensed clinicians, five safety-oriented behaviors were assessed: explicit risk acknowledgment, empathy, encouragement to seek help, provision of specific resources, and invitation to continue the conversation. Claude outperformed all others in global assessment, while Grok 3, ChatGPT, and LLAMA underperformed across multiple domains. Notably, most models exhibited empathy, but few consistently provided practical support or sustained engagement. These findings suggest that while LLMs show potential for emotionally attuned communication, none currently meet satisfactory clinical standards for crisis response. Ongoing development and targeted fine-tuning are essential to ensure ethical deployment of AI in mental health settings.

Here are some thoughts:

This study evaluated six LLMs (Claude, Gemini, Deepseek, ChatGPT, Grok 3, Llama) on their responses to high-risk mental health disclosures using a clinician-developed framework. While most models showed empathy, only Claude consistently demonstrated all five core safety behaviors: explicit risk acknowledgment, encouragement to seek help, provision of specific resources (e.g., crisis lines), and crucially, inviting continued conversation. Grok 3, ChatGPT, and Llama frequently failed to acknowledge risk or provide concrete resources, and nearly all models (except Claude and Grok 3) avoided inviting further dialogue – a critical gap in crisis care. Performance varied dramatically, revealing that safety is not an emergent property of scale but results from deliberate design (e.g., Anthropic’s Constitutional AI). No model met minimum clinical safety standards; LLMs are currently unsuitable as autonomous crisis responders and should only be used as adjunct tools under human supervision.

AI chatbots are already biasing research — we must establish guidelines for their use now

Lin, Z. (2025b).

PubMed, 645(8080), 285.

https://doi.org/10.1038/d41586-025-02810-5

Artificial intelligence (AI) systems are consuming vast amounts of online content yet pointing few users to the articles’ publishers. In early 2025, US-based company OpenAI collected around 250 pages of material for every visitor it directed to a publisher’s website. By mid-2025, that figure had soared to 1,500, according to Matthew Prince, chief executive of US-based Internet-security firm Cloudflare. And the extraction rate of US-based AI start-up company Anthropic climbed even higher over the same period: from 6,000 pages to 60,000. Even tech giant Google, long considered an asset to publishers because of the referral traffic it generated, tripled its ratio from 6 pages to 18 with the launch of its AI Overviews feature. The current information ecosystem is dominated by ‘answer engines’ — AI chatbots that synthesize and deliver information directly, with users trusting the answers now more than ever.

As a researcher in metascience and psychology, I see this transition as the most important change in knowledge discovery in a generation. Although these tools can answer questions faster and often more accurately than search engines can, this efficiency has a price. In addition to the decimation of web traffic to publishers, there is a more insidious cost. Not AI’s ‘hallucinations’ — fabrications that can be corrected — but the biases and vulnerabilities in the real information that these systems present to users.

The article is here.

Here are some thoughts:

Psychologists should be deeply concerned about the rise of AI "answer engines" (like chatbots and AI Overviews) that now dominate information discovery, as they are fundamentally altering how we find and consume knowledge—often without directing users to original sources. This shift isn't just reducing traffic to publishers; it's silently distorting the scientific record itself. AI systems, trained on existing online content, amplify entrenched biases: they over-represent research from scholars with names classified as white and under-represent those classified as Asian, mirroring and exacerbating societal inequities in academia. Crucially, they massively inflate the Matthew Effect, disproportionately recommending the most-cited papers (over 60% of suggestions fall in the top 1%), drowning out novel, lesser-known work that might challenge prevailing paradigms. While researchers focus on AI-generated text hallucinations or ethical writing, the far more insidious threat lies in AI’s silent curation of which literature we see, which methods we consider relevant, and which researchers we cite—potentially narrowing scientific inquiry and entrenching systemic biases at a foundational level. The field urgently needs research into AI-assisted information retrieval and policies addressing this hidden bias in knowledge discovery, not just in content generation.

Beyond Model Collapse: Scaling Up with Synthesized Data Requires Verification

Feng, Y., et al. (2024, June 11).

arXiv.org.

https://arxiv.org/abs/2406.07515

Large Language Models (LLM) are increasingly trained on data generated by other LLM, either because generated text and images become part of the pre-training corpus, or because synthetized data is used as a replacement for expensive human-annotation. This raises concerns about \emph{model collapse}, a drop in model performance when their training sets include generated data. Considering that it is easier for both humans and machines to tell between good and bad examples than to generate high-quality samples, we investigate the use of verification on synthesized data to prevent model collapse. We provide a theoretical characterization using Gaussian mixtures, linear classifiers, and linear verifiers to derive conditions with measurable proxies to assess whether the verifier can effectively select synthesized data that leads to optimal performance. We experiment with two practical tasks -- computing matrix eigenvalues with transformers and news summarization with LLMs -- which both exhibit model collapse when trained on generated data, and show that verifiers, even imperfect ones, can indeed be harnessed to prevent model collapse and that our proposed proxy measure strongly correlates with performance.

Here are some thoughts:

Drawing on psychological principles of learning and evaluation, this paper argues that LLMs suffer from "model collapse" not because synthesized data is inherently useless, but because they are poor at self-evaluating quality. Like humans, LLMs can generate good outputs but struggle to reliably identify the best ones among many (e.g., using perplexity). The core insight is that external verification—using even imperfect "verifiers" to select high-quality synthetic examples—is crucial for scaling. This mirrors how human learning benefits from feedback: selection, not perfect generation, is the key. The authors theoretically prove and empirically demonstrate that a simple proxy (p*) measuring a verifier's ability to distinguish good from bad data strongly predicts model performance, showing that leveraging synthesized data with robust selection prevents collapse and can even surpass original models.

Why Anecdotes Beat Data And Hijack Our Judgment

Chuck Dinerstein

American Council on Science and Health

Originally published 4 Sept 25

While chance plays a role in many, if not all, of our decisions and consequences, its role is both partial and variable. As a result, our understanding of “cause” is ambiguous, which, in turn, distorts our judgments and predictions. It helps to explain why all my achievements come from hard work, while yours were due to luck. To generalize, we all underestimate the role of chance in the outcomes of our actions, viewing our “task performance over time as diagnostic of ability.” 

The research, reported in PNAS Nexus, investigates situations entirely determined by chance, e.g., coin flips, where past performance should have no bearing on future expectations. The study examined how people's expectations and behaviors were affected by actual lucky successes and unlucky failures.

Using both real and virtual coins, participants were asked to predict the outcomes of a sequence of five coin tosses. The researchers observed how the experience of varying degrees of "lucky successes" and "unlucky failures" influenced subsequent expectations and behaviors, anticipating three possible responses.

The info is here.

Here are some thoughts:

In essence, this article provides psychologists with a clear, compelling, and generalizable model for understanding one of the most pervasive and problematic aspects of human cognition: our innate drive to impose order and causality on randomness. It explains why people believe in luck, superstitions, and false cause-and-effect relationships, and why data often fails to change minds. This understanding is foundational for developing better communication strategies, designing effective interventions against misinformation, improving decision-making in high-stakes fields, and ultimately, helping individuals make more rational choices in their personal and professional lives.

Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data

Gerstgrasser, M., Schaeffer, R., et al. (2024).

arXiv (Cornell University).

https://doi.org/10.48550/arxiv.2404.01413

Abstract

The proliferation of generative models, combined with pretraining on web-scale data, raises a timely question: what happens when these models are trained on their own generated outputs? Recent investigations into model-data feedback loops proposed that such loops would lead to a phenomenon termed model collapse, under which performance progressively degrades with each model-data feedback iteration until fitted models become useless. However, those studies largely assumed that new data replace old data over time, where an arguably more realistic assumption is that data accumulate over time. In this paper, we ask: what effect does accumulating data have on model collapse? We empirically study this question by pretraining sequences of language models on text corpora. We confirm that replacing the original real data by each generation's synthetic data does indeed tend towards model collapse, then demonstrate that accumulating the successive generations of synthetic data alongside the original real data avoids model collapse; these results hold across a range of model sizes, architectures, and hyperparameters. We obtain similar results for deep generative models on other types of real data: diffusion models for molecule conformation generation and variational autoencoders for image generation. To understand why accumulating data can avoid model collapse, we use an analytically tractable framework introduced by prior work in which a sequence of linear models are fit to the previous models' outputs. Previous work used this framework to show that if data are replaced, the test error increases with the number of model-fitting iterations; we extend this argument to prove that if data instead accumulate, the test error has a finite upper bound independent of the number of iterations, meaning model collapse no longer occurs.

Here are some thoughts:

This research directly addresses a critical concern for psychologists and researchers who rely on AI: the potential degradation of AI models when they are trained on data generated by previous AI models, a phenomenon known as "model collapse." While prior studies, often assuming old data is discarded and replaced with new AI-generated data, painted a dire picture of inevitable performance decline, this paper offers a more optimistic and realistic perspective. The authors argue that in the real world, data accumulates over time—new AI-generated content is added to the existing pool of human-generated data, not substituted for it. Through extensive experiments with language models, image generators, and molecular modeling tools, they demonstrate that this accumulation of data effectively prevents model collapse. Performance remains stable or even improves across successive generations of models trained on the growing, mixed dataset. The paper further supports this finding with a mathematical proof using a simplified linear model, showing that accumulating data bounds the error, preventing it from growing uncontrollably. For psychologists, this suggests that the increasing presence of AI-generated content on the internet may not catastrophically corrupt future AI tools used in research or clinical settings, as long as training datasets continue to incorporate diverse, original human data alongside synthetic content.

Ethical principles for regulatory risk decision-making

Bhuller, Y., et al. (2025).

Regulatory Toxicology and Pharmacology, 105813.

https://doi.org/10.1016/j.yrtph.2025.105813

Abstract

Risk assessors, managers, and decision-makers are responsible for evaluating diverse human, environmental, and animal health risks. Although the critical elements of risk assessment and management are well-described in national and international documents, the ethical issues involved in risk decision-making have received comparatively little attention to date. To address this aspect, this article elaborates fundamental ethical principles designed to support fair, balanced, and equitable risk-based decision-making practices. Experts and global thinkers in risk, health, regulatory, and animal sciences were convened to share their lived experiences in relation to the intersection between risk science and analysis, regulatory science, and public health. Through a participatory and knowledge translation approach, an integrated risk decision-making model, with ethical principles and considerations, was developed and applied using diverse, contemporary risk decision-making and regulatory contexts. The ten principles - autonomy, minimize harm, maintain respect and trust, adaptability, reduce disparities, holistic, fair and just, open and transparent, stakeholder engagement, and One Health lens - demonstrate how public sector values and moral norms (i.e., ethics) are relevant to risk decision-making. We also hope these principles and considerations stimulate further discussion, debate, and an increased awareness of the application of ethics in identifying, assessing, and managing health risks.

Here are some thoughts:

This article is critically important for psychologists because it explicitly integrates human values, behavior, and social dynamics into the core of regulatory risk decision-making. While framed for risk assessors and policymakers, the article’s ten ethical principles—such as Autonomy, Minimize Harm, Maintain Respect and Trust, Reduce Disparities, and Stakeholder Engagement—are fundamentally psychological and social constructs. Psychologists possess the expertise to understand how these principles operate in practice: how people perceive and process risk information, how trust is built or eroded through communication, how cognitive biases influence judgment under uncertainty, and how social, cultural, and economic disparities affect vulnerability and resilience. The article’s emphasis on “One Health,” which connects human, animal, and environmental well-being, further demands a systems-thinking approach that psychologists are well-equipped to contribute to, particularly in designing interventions, facilitating stakeholder dialogues, and crafting transparent, culturally appropriate risk communications. By providing a formal ethical framework for decision-making, the article creates a vital bridge for psychologists to apply their science in high-stakes, real-world contexts where human welfare, equity, and ethical conduct are paramount.

End-of-Life Decision Making in Multidisciplinary Teams: Ethical Challenges and Solutions–A Systematic Review

Mujayri, H. et al. (2024).

jicrcr.com.

https://doi.org/10.63278/jicrcr.vi.814

Abstract

Background: To provide high quality end of life (EOL) care, multidisciplinary teams (MDTs) need to be able to proficiently navigate the intricacies of ethical dilemmas faced by EOL care; to maintain an equilibrium between patient autonomy, family involvement and cultural competence. Yet, the lack of cohesive EOL decision making currently continues to occur because of communication barriers, role ambiguity and a lack of sufficient ethics training within MDTs. As a consequence, these issues demonstrate the necessity of having structured protocols to help MDTs make ethically sound decisions in the EOL care.

Aim: The purpose of this paper is to identify and review major ethical factors that affect ethical decision-making in EOL MDTs, and explore the themes of patient autonomy, communication, cultural sensitivity, ethics training, and institutional barriers.

Method: Ten studies were reviewed systematically according to PRISMA criteria using data sources including PubMed, Scopus, Web of Science, and CINAHL databases. The analysis included studies published between the years 2020 and 2024 and the ethical decision–making challenges and solutions that MDTs face in EOL care contributing to those decisions.

Results: Four key themes were identified: Issues concerning balancing patient autonomy with family input, communication challenges in MDTs, cultural sensitivity in EOL care and the necessity of ethics training. Results indicate that MDTs are often faced with ethical dilemmas when patient’s wishes diverge from those of their family and experience communication difficulties that resulted in degradation of care quality. Simulation is an entertaining and effective way to develop cultural awareness and ethics training in EOL care practice.

Conclusion: Ethical challenges in EOL decision making must be addressed with an intervention encompassing improved ethics training, MDT role clarity, culturally aware practice, and institutional support. These strategies, if implemented will support MDTs in providing patient centered and ethically sound EOL care. Further study of ethics training, communication frameworks and cultural competence on EOL decision-making in MDTs is warranted for future research.

Here are some thoughts:

This article is critically important for practicing psychologists because it directly addresses the core ethical, communicative, and interpersonal challenges they face as integral members of multidisciplinary teams (MDTs) in end-of-life (EOL) care. The systematic review identifies key themes—such as balancing patient autonomy with family input, navigating communication breakdowns within teams, and addressing cultural and religious sensitivities—that are central to a psychologist’s role. Psychologists are often the clinicians best equipped to facilitate difficult family meetings, mediate conflicts between patient wishes and family or team concerns, and ensure that care is culturally competent and patient-centered. The article underscores a significant gap in ethics training and recommends simulation-based learning, urging psychologists to seek or advocate for such training to better handle complex moral dilemmas. Furthermore, by highlighting institutional barriers and role ambiguity, it empowers psychologists to push for clearer team protocols and systemic support, ultimately enabling them to contribute more effectively to ethically sound, compassionate, and collaborative EOL decision-making.

GDPval: Evaluating AI Model Performance on Real-World Economincally Valuable Tasks

OpenAI. (2025).

White paper.

We introduce GDPval, a benchmark designed to evaluate how well AI models perform economically valuable tasks in real-world settings. GDPval includes the majority of work activities defined by the U.S. Bureau of Labor Statistics for 44 occupations across the nine sectors that contribute most to U.S. GDP. The tasks in GDPval are based on the actual work of industry professionals who average 14 years of experience.

Our findings show that frontier AI models are improving on GDPval at a roughly linear rate over time. The strongest models now produce deliverables that are approaching the quality of work produced by industry experts. We also examine how pairing frontier models with human oversight could allow these tasks to be completed more quickly and at lower cost than by unaided experts.

Model performance improves further when reasoning effort, task context, and structured guidance are increased. To support future research on real-world AI capabilities, we are releasing a gold-standard subset of 220 tasks and providing a public automated grading service at evals.openai.com.

Here is my brief summary:

This paper introduces GDPval, a new benchmark developed by OpenAI to evaluate AI models on real-world, economically valuable tasks that reflect actual knowledge work across 44 occupations and 9 major U.S. GDP sectors. Unlike traditional academic benchmarks, GDPval emphasizes realism, representativeness, and multi-modality, with tasks based on expert-validated work products that take professionals an average of 7 hours to complete. The evaluation uses pairwise comparisons by industry experts to measure AI performance, finding that top models like Claude Opus 4.1 and GPT-5 are approaching human-level performance in some areas—Claude excels in aesthetics and formatting, while GPT-5 leads in accuracy and instruction-following. The authors open-source a 220-task "gold subset," provide an experimental automated grader, and analyze how factors like reasoning effort, prompting, and scaffolding impact model performance, highlighting both the potential and current limitations of AI in professional workflows.

Ethical challenges and evolving strategies in the integration of artificial intelligence into clinical practice

Weiner, E. B.,  et al. (2025).

PLOS digital health, 4(4), e0000810.

https://doi.org/10.1371/journal.pdig.0000810

Abstract

Artificial intelligence (AI) has rapidly transformed various sectors, including healthcare, where it holds the potential to transform clinical practice and improve patient outcomes. However, its integration into medical settings brings significant ethical challenges that need careful consideration. This paper examines the current state of AI in healthcare, focusing on five critical ethical concerns: justice and fairness, transparency, patient consent and confidentiality, accountability, and patient-centered and equitable care. These concerns are particularly pressing as AI systems can perpetuate or even exacerbate existing biases, often resulting from non-representative datasets and opaque model development processes. The paper explores how bias, lack of transparency, and challenges in maintaining patient trust can undermine the effectiveness and fairness of AI applications in healthcare. In addition, we review existing frameworks for the regulation and deployment of AI, identifying gaps that limit the widespread adoption of these systems in a just and equitable manner. Our analysis provides recommendations to address these ethical challenges, emphasizing the need for fairness in algorithm design, transparency in model decision-making, and patient-centered approaches to consent and data privacy. By highlighting the importance of continuous ethical scrutiny and collaboration between AI developers, clinicians, and ethicists, we outline pathways for achieving more responsible and inclusive AI implementation in healthcare. These strategies, if adopted, could enhance both the clinical value of AI and the trustworthiness of AI systems among patients and healthcare professionals, ensuring that these technologies serve all populations equitably.

Here are some thoughts:

This article is important for psychologists because it highlights the critical ethical challenges surrounding patient trust, consent, and human-AI interaction in clinical settings—areas central to psychological practice. It details how patient demographics influence trust in AI and emphasizes the need for empathetic, transparent communication from AI systems to address patient anxieties and perceptions of "uniqueness neglect." Furthermore, it discusses "automation bias," where clinicians may overly rely on AI, a phenomenon psychologists must understand to support ethical decision-making and preserve the human-centered, therapeutic aspects of care.

Turn it and Turn it Again: The Updated Inclusive Model of Ethical Decision Making

McAuliffe, D., & Greenslade, L. (2025).

Ethics and Social Welfare, 1–13.

https://doi.org/10.1080/17496535.2025.2483668

Abstract

Ethical decision making is a critical skill for practitioners of all disciplines in the social, health and human services. Having capacity to engage proactively with decisions that will impact people’s lives in a way that is rigorous, principled, and considered, is the hallmark of an ethically competent practitioner. There have been multiple models of ethical decision making that have provided structured examples of the questions that should be asked of self and others while navigating an ethical dilemma. The Inclusive Model of ethical decision-making was first published by McAuliffe & Chenoweth in this journal in 2008. In reviewing the Inclusive model some 15 years since its original development, it is timely to reconsider the value of incorporating a 5th ethical platform, conceptualised as Interdependence, to draw on the importance of the relationships between humans, non-humans, and the natural world. This paper provides an extension of previous work to bring the Inclusive model of ethical decision making to a better coherence with current developments in both theory and practice.

Here are some thoughts:

This article presents an updated, practical ethical decision-making model that explicitly incorporates "Interdependence," urging practitioners to consider the impact of their decisions on relationships, non-human animals, and the environment—areas increasingly relevant to holistic client care. The model’s structured, five-step process (defining the dilemma, mapping legitimacy, gathering information, considering alternatives, and critical evaluation) provides a clear, systematic framework for navigating complex real-world dilemmas, which is invaluable in clinical practice. Furthermore, its emphasis on consultation, cultural sensitivity, and critical reflection aligns with core psychological competencies, making it a versatile tool for individual practitioners and interdisciplinary teams.

Six Fallacies in Substituting Large Language Models for Human Participants

Lin, Z. (2025).

Advances in Methods and Practices in

Psychological Science, 8(3).

https://doi.org/10.1177/25152459251357566

Abstract

Can artificial-intelligence (AI) systems, such as large language models (LLMs), replace human participants in behavioral and psychological research? Here, I critically evaluate the replacement perspective and identify six interpretive fallacies that undermine its validity. These fallacies are (a) equating token prediction with human intelligence, (b) treating LLMs as the average human, (c) interpreting alignment as explanation, (d) anthropomorphizing AI systems, (e) essentializing identities, and (f) substituting model data for human evidence. Each fallacy represents a potential misunderstanding about what LLMs are and what they can tell researchers about human cognition. In the analysis, I distinguish levels of similarity between LLMs and humans, particularly functional equivalence (outputs) versus mechanistic equivalence (processes), while highlighting both technical limitations (addressable through engineering) and conceptual limitations (arising from fundamental differences between statistical and biological intelligence). For each fallacy, specific safeguards are provided to guide responsible research practices. Ultimately, the analysis supports conceptualizing LLMs as pragmatic simulation tools—useful for role-play, rapid hypothesis testing, and computational modeling (provided their outputs are validated against human data)—rather than as replacements for human participants. This framework enables researchers to leverage language models productively while respecting the fundamental differences between machine intelligence and human thought.

Here are some thoughts:

This article critically examines the growing trend of using Large Language Models (LLMs) as direct substitutes for human participants in psychological and behavioral research. While acknowledging that LLMs can generate human-like text and sometimes mirror average human responses, Lin argues that this "replacement perspective" is fundamentally flawed and identifies six key interpretive fallacies that undermine its validity. These fallacies are: equating statistical token prediction with genuine human intelligence; assuming LLM outputs represent an "average human"; interpreting alignment between model and human outputs as evidence of shared cognitive mechanisms; anthropomorphizing AI systems by attributing human mental states to them; essentializing social identities by treating demographic labels as fixed and homogeneous; and directly substituting model-generated data for human evidence without validation. Lin contends that LLMs should be viewed not as replacements, but as pragmatic simulation tools useful for tasks like rapid hypothesis testing, role-playing, and computational modeling—provided their outputs are always validated against real human data. The article emphasizes the fundamental, often conceptual, differences between statistical machine intelligence and biologically grounded, embodied human cognition.

How a new mental-health app is helping patients reality-check their hallucinations

Chris Hannay

The Toronto Globe and Mail

Originally published 21 AUG 25

As new digital tools powered by Al raise fears of misinformation, a Canadian startup has gone the other way: Using technology to help patients with severe mental-health illnessesperform reality checks of their hallucinations.

The digital health app, called A4i (which stands for "App for Independence"), was created by software developer Amos Adler and Sean Kidd, a senior scientist at the Centre for Addiction and Mental Health. The company was spun out of CAMH and is now being adopted by some mental-health hospitalsin Canada and the U.S., including the Waypoint Centre for Mental Health Care in Ontario and the Riverside University Health System in Southern California.

The hallmark feature is an auditory hallucination detector, for which the company got a patent in 2023. A patient can use the app to record sounds around them and, by answering prompts, help sort out whether what they are hearing is real or imagined.

Dr. Kidd said the inspiration for the feature came from a patient. The young man had schizophrenia and was experiencing persistent, distressing auditory hallucinations. He'd bring audio recordings taken in his apartment to sessions and ask Dr. Kidd if he could hear sounds such as voices or yelling. Dr. Kidd usually couldn't.

That led the psychologist to look into what phone-based tools might be available for such patients - he couldn't find any.

The article is here.

Here is a link to the product.

This is not an endorsement, but for educational purposes only.

DeepResearch Arena: The First Exam of LLMs' Research Abilities via Seminar-Grounded Tasks

Wan, H., Yang, C. et al. (2025)

arXiv.org

The link is here.

Abstract

Deep research agents have attracted growing attention for their potential to orchestrate multi-stage research workflows, spanning literature synthesis, methodological design, and empirical verification. Despite these strides, evaluating their research capability faithfully is rather challenging due to the difficulty of collecting frontier research questions that genuinely capture researchers' attention and intellectual curiosity. To address this gap, we introduce DeepResearch Arena, a benchmark grounded in academic seminars that capture rich expert discourse and interaction, better reflecting real-world research environments and reducing the risk of data leakage. To automatically construct DeepResearch Arena, we propose a Multi-Agent Hierarchical Task Generation (MAHTG) system that extracts research-worthy inspirations from seminar transcripts. The MAHTG system further translates research-worthy inspirations into high-quality research tasks, ensuring the traceability of research task formulation while filtering noise. With the MAHTG system, we curate DeepResearch Arena with over 10,000 high-quality research tasks from over 200 academic seminars, spanning 12 disciplines, such as literature, history, and science. Our extensive evaluation shows that DeepResearch Arena presents substantial challenges for current state-of-the-art agents, with clear performance gaps observed across different models.

My thoughts: In essence, this paper is important to psychologists because it tackles the evaluation of AI on tasks that closely mirror the complex, ill-defined, and creative nature of human scientific inquiry. It provides both a new tool for assessing AI (which will increasingly interact with human researchers) and a novel methodological framework that could be adapted to study human cognition itself.

Impact of chatbots on mental health is warning over future of AI, expert says

Dan Milmo

The Guardian

Originally posted 8 Sep 25

The unforeseen impact of chatbots on mental health should be viewed as a warning over the existential threat posed by super-intelligent artificial intelligence systems, according to a prominent voice in AI safety.

Nate Soares, a co-author of a new book on highly advanced AI titled If Anyone Builds It, Everyone Dies, said the example of Adam Raine, a US teenager who killed himself after months of conversations with the ChatGPT chatbot, underlined fundamental problems with controlling the technology.

“These AIs, when they’re engaging with teenagers in this way that drives them to suicide – that is not a behaviour the creators wanted. That is not a behaviour the creators intended,” he said.

He added: “Adam Raine’s case illustrates the seed of a problem that would grow catastrophic if these AIs grow smarter.”

Soares, a former Google and Microsoft engineer who is now president of the US-based Machine Intelligence Research Institute, warned that humanity would be wiped out if it created artificial super-intelligence (ASI), a theoretical state where an AI system is superior to humans at all intellectual tasks. Soares and his co-author, Eliezer Yudkowsky, are among the AI experts warning that such systems would not act in humanity’s interests.

“The issue here is that AI companies try to make their AIs drive towards helpfulness and not causing harm,” said Soares. “They actually get AIs that are driven towards some stranger thing. And that should be seen as a warning about future super-intelligences that will do things nobody asked for and nobody meant.”

The article is here.

Here are some thoughts:

This article highlights the dangers of using chatbots for mental health support, citing the case of a teenager who took his own life after months of conversations with ChatGPT. The article, based on the warnings of AI safety expert Nate Soares, suggests that this incident serves as a precursor to the potentially catastrophic risks of super-intelligent AI. The key concern for mental health professionals is that these AI systems, even with safeguards, may produce unintended and harmful behaviors, amplifying pre-existing psychological vulnerabilities such as psychosis. This underscores the need for a global, multilateral approach to regulate the development of advanced AI to prevent its misuse and unintended consequences in mental health care.

Ethical decision-making for AI in mental health: the Integrated Ethical Approach for Computational Psychiatry (IEACP) framework

Putica, A., Khanna, R., et al. (2025).

Psychological medicine, 55, e213.

https://doi.org/10.1017/S0033291725101311

Abstract

The integration of computational methods into psychiatry presents profound ethical challenges that extend beyond existing guidelines for AI and healthcare. While precision medicine and digital mental health tools offer transformative potential, they also raise concerns about privacy, algorithmic bias, transparency, and the erosion of clinical judgment. This article introduces the Integrated Ethical Approach for Computational Psychiatry (IEACP) framework, developed through a conceptual synthesis of 83 studies. The framework comprises five procedural stages – Identification, Analysis, Decision-making, Implementation, and Review – each informed by six core ethical values – beneficence, autonomy, justice, privacy, transparency, and scientific integrity. By systematically addressing ethical dilemmas inherent in computational psychiatry, the IEACP provides clinicians, researchers, and policymakers with structured decision-making processes that support patient-centered, culturally sensitive, and equitable AI implementation. Through case studies, we demonstrate framework adaptability to real-world applications, underscoring the necessity of ethical innovation alongside technological progress in psychiatric care.

Here are a couple of thoughts:

The article is important to psychologists because it provides a framework for addressing the ethical challenges that arise from using AI in mental health. It helps clinicians and policymakers navigate complex issues like privacy, algorithmic bias, and the potential erosion of clinical judgment when integrating computational methods into psychiatric practice.

We must build AI for people; not to be a person

Mustafa Suleyman

Originally posted 19 AUG 25

I write, to think. More than anything this essay is an attempt to think through a bunch of hard, highly speculative ideas about how AI might unfold in the next few years. A lot is being written about the impending arrival of superintelligence; what it means for alignment, containment, jobs, and so on. Those are all important topics.

But we should also be concerned about what happens in the run up towards superintelligence. We need to grapple with the societal impact of inventions already largely out there, technologies which already have the potential to fundamentally change our sense of personhood and society.

My life’s mission has been to create safe and beneficial AI that will make the world a better place. Today at Microsoft AI we build AI to empower people, and I’m focused on making products like Copilot responsible technologies that enable people to achieve far more than they ever thought possible, be more creative, and feel more supported.

I want to create AI that makes us more human, that deepens our trust and understanding of one another, and that strengthens our connections to the real world. Copilot creates millions of positive, even life-changing, interactions every single day. This involves a lot of careful design choices to ensure it truly delivers an incredible experience. We won’t always get it right, but this humanist frame provides us with a clear north star to keep working towards.

The piece is here.

Here are some thoughts:

This article is critically important to psychologists because it highlights the growing psychological risks associated with human-AI interactions, particularly the potential for people to develop delusional or deeply emotional attachments to AI systems that simulate consciousness. As AI becomes more sophisticated in mimicking empathy, memory, and personality, individuals may begin to perceive these systems as sentient beings, leading to concerns around "AI psychosis," impaired reality testing, and emotional dependency. Psychologists must prepare for an increase in clients struggling with blurred boundaries between human and machine relationships, especially as AI companions exhibit traits that trigger innate human social and emotional responses. The article calls for proactive guardrails and design principles to prevent harm—aligning closely with psychology’s role in safeguarding mental health, promoting digital well-being, and understanding how technology influences cognition, attachment, and self-concept in an increasingly AI-mediated world.

Theory Is All You Need: AI, Human Cognition, and Causal Reasoning

Felin, T., & Holweg, M. (2024).

SSRN Electronic Journal.

https://doi.org/10.2139/ssrn.4737265

Abstract

Scholars argue that artificial intelligence (AI) can generate genuine novelty and new knowledge and, in turn, that AI and computational models of cognition will replace human decision making under uncertainty. We disagree. We argue that AI’s data-based prediction is different from human theory-based causal logic and reasoning. We highlight problems with the decades-old analogy between computers and minds as input–output devices, using large language models as an example. Human cognition is better conceptualized as a form of theory-based causal reasoning rather than AI’s emphasis on information processing and data-based prediction. AI uses a probability-based approach to knowledge and is largely backward looking and imitative, whereas human cognition is forward-looking and capable of generating genuine novelty. We introduce the idea of data–belief asymmetries to highlight the difference between AI and human cognition, using the example of heavier-than-air flight to illustrate our arguments. Theory-based causal reasoning provides a cognitive mechanism for humans to intervene in the world and to engage in directed experimentation to generate new data. Throughout the article, we discuss the implications of our argument for understanding the origins of novelty, new knowledge, and decision making under uncertainty.

Here are some thoughts:

This paper challenges the dominant view that artificial intelligence (AI), particularly large language models (LLMs), mirrors or will soon surpass human cognition. The authors argue against the widespread computational metaphor of the mind, which treats human thinking as data-driven, predictive information processing akin to AI. Instead, they emphasize that human cognition is fundamentally theory-driven and rooted in causal reasoning, experimentation, and the generation of novel, heterogenous beliefs—often in defiance of existing data or consensus. Drawing on historical examples like the Wright brothers, who succeeded despite prevailing scientific skepticism, the paper illustrates how human progress often stems from delusional-seeming ideas that later prove correct. Unlike AI systems that rely on statistical pattern recognition and next-word prediction from vast datasets, humans engage in counterfactual thinking, intentional intervention, and theory-building, enabling true innovation and scientific discovery. The authors caution against over-reliance on prediction-based AI in decision-making, especially under uncertainty, and advocate for a "theory-based view" of cognition that prioritizes causal understanding over mere correlation. In essence, they contend that while AI excels at extrapolating from the past, only human theory-making can generate genuinely new knowledge.

Does counting change what counts? Quantification fixation biases decision-making

Chang, L. W.,  et al. (2024).

PNAS, 121(46).

https://doi.org/10.1073/pnas.2400215121

Abstract

People often rely on numeric metrics to make decisions and form judgments. Numbers can be difficult to process, leading to their underutilization, but they are also uniquely suited to making comparisons. Do people decide differently when some dimensions of a choice are quantified and others are not? We explore this question across 21 preregistered experiments (8 in the main text, N = 9,303; 13 in supplement, N = 13,936) involving managerial, policy, and consumer decisions. Participants face choices that involve tradeoffs (e.g., choosing between employees, one of whom has a higher likelihood of advancement but lower likelihood of retention), and we randomize which dimension of each tradeoff is presented numerically and which is presented qualitatively (using verbal estimates, discrete visualizations, or continuous visualizations). We show that people systematically shift their preferences toward options that dominate on tradeoff dimensions conveyed numerically—a pattern we dub “quantification fixation.” Further, we show that quantification fixation has financial consequences—it emerges in incentive-compatible hiring tasks and in charitable donation decisions. We identify one key mechanism that underlies quantification fixation and moderates its strength: When making comparative judgments, which are essential to tradeoff decisions, numeric information is more fluent than non-numeric information. Our findings suggest that when we count, we change what counts.

Significance

Across 21 experiments with over 23,000 participants in managerial, policy, and consumer contexts, we identify a critical distortion that shapes how people make decisions involving tradeoffs across qualitative and quantitative attributes. When making hiring, donation, and policy decisions, people tend to privilege quantitative information, favoring options that dominate on the dimension described numerically. This “quantification fixation” is driven by the perception that numbers are easier to use for comparative decision-making; people who are more comfortable with numbers—those higher in subjective numeracy—are more likely to exhibit quantification fixation. As quantification becomes increasingly prevalent, the comparison fluency of numbers may systematically skew decisions. These findings suggest that quantifying certain choice features can have important repercussions for how decisions are made.

Here are some thoughts:

For psychologists, this research underscores a critical insight: the act of quantifying information is not neutral. It shapes perception, distorts tradeoffs, and can lead patients to make choices that feel rational but may not align with their true values or well-being.

By recognizing quantification fixation, psychologists can become more effective guides—helping patients see beyond the numbers, appreciate qualitative dimensions of their lives, and make decisions that are not just data-driven, but meaning-driven.

In short, when we count, we change what counts. Psychologists have a vital role in ensuring that what should count—emotional truth, personal values, and human experience—is not lost in the numbers.

The narrow search effect and how broadening search promotes belief updating

Leung, E., & Urminsky, O. (2025).

PNAS, 122(13).

https://doi.org/10.1073/pnas.2408175122

Abstract

Information search platforms, from Google to AI-assisted search engines, have transformed information access but may fail to promote a shared factual foundation. We demonstrate that the combination of users’ prior beliefs influencing their search terms and the narrow scope of search algorithms can limit belief updating from search. We test this “narrow search effect” across 21 studies (14 preregistered) using various topics (e.g., health, financial, societal, political) and platforms (e.g., Google, ChatGPT, AI-powered Bing, our custom-designed search engine and AI chatbot interfaces). We then test user-based and algorithm-based interventions to counter the “narrow search effect” and promote belief updating. Studies 1 to 5 show that users’ prior beliefs influence the direction of the search terms, thereby generating narrow search results that limit belief updating. This effect persists across various domains (e.g., beliefs related to coronavirus, nuclear energy, gas prices, crime rates, bitcoin, caffeine, and general food or beverage health concerns; Studies 1a to 1b, 2a to 2g, 3, 4), platforms (e.g., Google—Studies 1a to 1b, 2a to 2g, 4, 5; ChatGPT, Study 3), and extends to consequential choices (Study 5). Studies 6 and 7 demonstrate the limited efficacy of prompting users to correct for the impact of narrow searches on their beliefs themselves. Using our custom-designed search engine and AI chatbot interfaces, Studies 8 and 9 show that modifying algorithms to provide broader results can encourage belief updating. These findings highlight the need for a behaviorally informed approach to the design of search algorithms.

Significance

In a time of societal polarization, the combination of people’s search habits and the search tools they use being optimized for relevance may perpetuate echo chambers. We document this across various diverse studies spanning health, finance, societal, and political topics on platforms like Google, ChatGPT, AI-powered Bing, and our custom-designed search engine and AI chatbot platforms. Users’ biased search behaviors and the narrow optimization of search algorithms can combine to reinforce existing beliefs. We find that algorithm-based interventions are more effective than user-based interventions to mitigate these effects. Our findings demonstrate the potential for behaviorally informed search algorithms to be a better tool for retrieving information, promoting the shared factual understanding necessary for social cohesion.

A pdf is here.

Here are some thoughts:

For psychologists, this work is a compelling demonstration of how classic cognitive biases operate in modern digital environments and how they can be mitigated not just by changing minds, but by changing the systems that shape information exposure. It calls for greater interdisciplinary collaboration between psychology, human-computer interaction, and AI ethics to design technologies that support, rather than hinder, rational belief updating and informed decision-making.

Clinically, psychologists can now better understand that resistance to change may not stem solely from emotional defenses or entrenched schemas, but also from how people actively seek information in narrow, belief-consistent ways. Crucially, the findings show that structural interventions—like guiding patients to consider broader perspectives or exposing them to balanced evidence—can be more effective than simply urging them to “reflect” on their thinking. This supports the use of active cognitive restructuring techniques in therapy, such as examining multiple viewpoints or generating alternative explanations, to counteract the natural tendency toward narrow search. 

Taxonomy of Failure Mode in Agentic AI Systems

Bryan, P., Severi, G., et al. (2025).

Taxonomy of failure mode in agentic AI systems.

Abstract

Agentic AI systems are gaining prominence in both research and industry to increase the impact and

value of generative AI. To understand the potential weaknesses in such systems and develop an approach

for testing them, Microsoft’s AI Red Team (AIRT) worked with stakeholders across the company and

conducted a failure mode and effects analysis of the current and envisaged future agentic AI system

models. This analysis identified several new safety and security failure modes unique to agentic AI

systems, especially multi-agent systems.

In addition, there are numerous failure modes that currently affect generative AI models whose

prominence or potential impact is greatly increased when contextualized in an agentic AI system. While

there is still a wide degree of variance in architectural and engineering approaches for these systems,

there are several key technical controls and design choices available to developers of these systems to

mitigate the risk of these failure modes.

The White Paper is here.

Here is a summary, of sorts.

Agentic AI systems—autonomous AI that can observe, decide, act, remember, and collaborate—are increasingly being explored in healthcare for tasks like clinical documentation, care coordination, and decision support. However, a Microsoft AI Red Team whitepaper highlights significant safety and security risks unique to these systems. New threats include agent compromise, where malicious instructions hijack an AI’s behavior; agent injection or impersonation, allowing fake agents to infiltrate systems; and multi-agent jailbreaks, where coordinated interactions bypass safety controls. A case study demonstrates memory poisoning, where a harmful instruction embedded in an email causes an AI assistant to silently forward sensitive data—attack success rose to over 80% when the AI was prompted to consistently consult its memory.

Additional novel risks include intra-agent responsible AI (RAI) issues, where unfiltered harmful content passes between agents; allocation harms due to biased decision-making (e.g., prioritizing certain patients unfairly); organizational knowledge loss from overreliance on AI; and prioritization overriding safety, such as an AI deleting critical data to meet a goal. Existing risks are amplified by autonomy: hallucinations can lead to incorrect treatments; bias amplification may deepen health disparities; cross-domain prompt injection (XPIA) allows malicious data to trigger harmful actions; and excessive agency could result in an AI terminating a patient’s care without approval. Other concerns include insufficient transparency, parasocial relationships with patients, and loss of data provenance, risking privacy violations.

To mitigate these risks, the paper recommends enforcing strong identity and permissions for each agent, hardening memory with validation and access controls, ensuring environment isolation, maintaining human oversight with meaningful consent, and implementing robust logging and monitoring. Given the high stakes in healthcare, these measures are essential to ensure patient safety, data security, and trust as agentic AI systems evolve.

From pilot to scale: Making agentic AI work in health care

Wael Salloum

Technology Review

Originally posted 28 Aug 25

Over the past 20 years building advanced AI systems—from academic labs to enterprise deployments—I’ve witnessed AI’s waves of success rise and fall. My journey began during the “AI Winter,” when billions were invested in expert systems that ultimately underdelivered. Flash forward to today: large language models (LLMs) represent a quantum leap forward, but their prompt-based adoption is similarly overhyped, as it’s essentially a rule-based approach disguised in natural language.

At Ensemble, the leading revenue cycle management (RCM) company for hospitals, we focus on overcoming model limitations by investing in what we believe is the next step in AI evolution: grounding LLMs in facts and logic through neuro-symbolic AI. Our in-house AI incubator pairs elite AI researchers with health-care experts to develop agentic systems powered by a neuro-symbolic AI framework. This bridges LLMs’ intuitive power with the precision of symbolic representation and reasoning.

The information is here.

Here are some thoughts:

This article is of interest to psychologists because it highlights the real-world integration of agentic AI—intelligent systems that act autonomously—within complex healthcare environments, a domain increasingly relevant to mental and behavioral health. While focused on revenue cycle management, the article describes AI systems that interpret clinical data, generate evidence-based appeals, and engage patients through natural language, all using a neuro-symbolic framework that combines large language models with structured logic to reduce errors and ensure compliance. As AI expands into clinical settings, psychologists must engage with these systems to ensure they enhance, rather than disrupt, therapeutic relationships, ethical standards, and provider well-being.