Cognitive processes in technology trust development

As AI becomes more embedded in how we work and make decisions, an important question is emerging: Are we becoming more efficient or are we quietly changing how we think? I’m pleased to share my latest peer-reviewed paper: “Beyond Efficiency: A Qualitative Exploration of Human Agency, Epistemic Vigilance, and Cognitive Boundaries in Human–AI Interaction.” This study was co-authored with Dr Jonathan Ee, PhD, a clinical psychologist, researcher and academic leader, drawing on qualitative interviews with professionals across psychology, technology and leadership. This work builds on my earlier research on Human-Centred AI design developed during my MSc in Psychology, moving from how we design AI to how people experience and trust it and now to how AI is reshaping human cognition and agency. Some key insights from the research: - AI significantly reduces cognitive load and accelerates tasks but shifts effort away from thinking towards monitoring and verification, with cognitive boundaries becoming more fluid as individuals negotiate what to think through themselves and what to delegate - Human agency must be actively maintained through deliberate judgement and oversight - Epistemic vigilance, the ability to question and scrutinise information, becomes critical when interacting with AI outputs - There is a growing risk that overreliance on AI may weaken creativity, critical thinking and sense of ownership over decisions A central finding is that AI does not replace human cognition. It reshapes it. This creates a new responsibility. Not just to use AI effectively but to remain the primary thinker in the process. For organisations, this has clear implications for workflows, governance and training. Efficiency alone cannot be the goal. We must also preserve human judgement and cognitive capability. Full paper here: https://lnkd.in/gf4VM_fC My other publications are available at www.techsynapse.ai I’d be interested in your perspective: How do you decide what to rely on AI for and what you need to think through yourself? #AI #HumanAIInteraction #HumanCentredAI #Psychology #CognitiveScience #ResponsibleAI #TrustInAI #AIethics #BuildingbetterAI #TechSynapse.ai

🔬 Paper Alert: Trust in AI is not built in isolation – it’s social. 🤖 Proud supervisor moment: My (and Simon B. de Jong´s) doctoral student Türkü Erengin has just published her very first paper, "You, Me, and the AI: The Role of Third-Party Human Teammates for Trust Formation Toward AI Teammates," in Wiley´s Journal of Organizational Behavior. 🤖 So, what does this research tell us? AI teammates are becoming a reality in modern workplaces. But while research has focused on how humans individually evaluate AI, Türkü’s work brings a fresh perspective: trust in AI is shaped severely by and learned from the people around us. Using two main (+ two supplementary) studies including a really cool observational, incentivized study with human-AI teams (including real GPT-powered physical service robot Temi, see picture)—this paper shows that: ✅ If a human teammate trusts an AI, their colleagues are more likely to trust it too. This effect is not only quite strong, it is also stable when controlling for people´s own, initial preferences after trying the AI the first time and holds true in contexts where actual money is on the table! ✅ This effect disappears if the human teammate themselves is seen as untrustworthy. ✅ Trust in AI is not just about the AI's own reliability—it depends on social context and human relationships. 🚀 Why does this matter? 1️⃣ Organizations implementing AI should focus on social dynamics and context rather than just AI performance. It does not (only) matter how well AI functions - if relevant others around employees don´t trust AI, employees won´t either. 2️⃣ Building trust in AI requires trusted human advocates—if key employees are skeptical, adoption suffers. 3️⃣ AI trust calibration is crucial: Over-reliance and under-reliance on AI both have risks, and leaders should consider social influences when introducing AI teammates. 🎉 Huge congratulations to Türkü for this important contribution! If you’re interested in how social cognitive theory can explain trust in AI teams, check out the full paper. What makes this even more special? JOB is the journal where my first academic paper was published—and where my own PhD supervisor (Frank Walter) had their first journal publication. A true academic full-circle moment! 🎓🔁 I’d love to hear from others: Have you noticed social influences shaping how people trust AI in your workplace? Have you ever seen CEOs, leaders, colleagues modeling (or refraining from modeling) trust in AI? #AI #TrustInAI #HumanAITeams #OrganizationalBehavior #FutureOfWork #Leadership #AcademicMentorship

Trust in technology is not about making systems look friendly or adding more explanations. It is about how people decide to rely on something when there is uncertainty. In human computer interaction, trust is a judgment users make. It is shaped by expectations, experience, social cues, perceived control, and context. The same system can be trusted in one situation and distrusted in another. That is why trust is so hard to design and so easy to break. Research shows that users do not trust systems for a single reason. Sometimes trust comes from reasoning. Does this system behave consistently? Does it do what I expect? Other times trust comes from feeling. Does this interface feel human, present, or socially responsive? In many cases trust is social. If people I trust rely on this system, I am more likely to trust it too. There are also moments where trust collapses. When users feel forced, manipulated, or stripped of control, distrust appears even if the system is accurate. When early experiences violate expectations, trust erodes fast and rarely recovers on its own. One of the most important insights is that trust is dynamic. It builds slowly through repeated positive interactions and can disappear quickly after a single negative one. Designing for trust is not about maximizing trust. It is about supporting appropriate trust. Helping users know when to rely on a system and when not to. For AI, automation, and complex digital products, this matters more than ever. Overtrust is just as dangerous as distrust. Good design respects user agency, supports understanding, and stays honest about limitations. Trust is not a feature you add at the end. It is an outcome of how the entire system behaves over time.

🤔 A fascinating new study from Oregon State University, GitHub, and Northern Arizona University researchers reveals what really drives developers to trust and adopt AI tools - and it's not what most of us assumed. As someone who's spent years studying organizational psychology and now helps companies navigate AI adoption initiatives, what caught my attention wasn't just what influences trust in AI - but what doesn't. Here are three surprising insights that challenge conventional wisdom: 1. Ease of use? Not a significant factor. Unlike traditional tech adoption, developers' trust in AI tools isn't heavily influenced by how easy they are to use. This suggests the game has changed - we're moving beyond basic usability concerns to deeper questions of value and alignment. 2. Trust is built on three pillars: - System/output quality (does it do what it claims?) - Functional value (does it provide tangible benefits?) - Goal maintenance (does it align with developers' objectives?) 3. 🔍 Most fascinating: Cognitive styles matter more than we thought. Developers who: - Are intrinsically motivated by technology - Have higher computer self-efficacy - Show greater risk tolerance  ...are significantly more likely to adopt these tools. Through my work at Fractional Insights, I've observed how organizations often focus on technical training while overlooking these psychological factors. But this research suggests we need a more nuanced approach to AI adoption - one that accounts for cognitive diversity and individual differences in how people approach new technology. 💡 The key takeaway for organizational leaders: Successful AI adoption isn't just about the technology - it's about understanding and supporting the diverse ways people think about and interact with these tools. What's your experience? What have you noticed about how psychology impacts AI tool adoption in your organization? Throwback pic to talking about technology and humanity with some of my favorite experts: Amir Ghowsi Moritz Sudhof at NYU with Anna A. Tavis, PhD. #FutureOfWork #OrganizationalPsychology #AIAdoption #TechnologyTransformation #InclusiveDesign #LeadershipInsights

Is AI Replacing Our Thinking? | ¿Estamos empezando a delegarle el pensamiento a la IA? Two recent research papers (links below) challenged my assumption that AI is simply a tool. The reality may be more complex...and these studies suggest a deeper shift may already be underway. 😱 AI is starting to reshape how we reason 😱 Let’s unpack this step by step. Keep reading! 👇 👇 👇 1️⃣ A new participant in human cognition For decades, behavioral science (popularized by Daniel Kahneman) described thinking as two systems: • System 1 (fast, intuitive, automatic) • System 2 (slow, analytical, effortful) But generative AI introduces a third actor: • Artificial cognition —> external reasoning generated by AI systems. 🤭 For the first time in history, part of our thinking process can occur outside the human mind. 2️⃣ The risk of "cognitive surrender" It happens when people rely heavily on AI outputs without critical evaluation. This doesn’t necessarily happen because people are careless. It happens because AI often appears confident, coherent, and fast...qualities that naturally trigger trust. 3️⃣ What real-world conversations reveal Studies provide rare empirical evidence from real usage: severe "disempowerment" cases were rare, but not negligible at scale. Researchers identified patterns where AI can subtly reduce human agency, including: • Users asking the AI to decide what they should do • Delegating personal communication (messages, decisions, judgments) • Treating the model as an authority rather than a collaborator Perhaps the most surprising finding: Interactions with higher disempowerment potential often receive higher user satisfaction scores. 🤯 🔥 4️⃣ The paradox of the AI era AI can dramatically improve decisions when used well. But if we stop questioning outputs, the outcome of many decisions becomes dependent on the AI rather than the human judgment behind it. 🫣 What does this mean? It seems clear to me: the real challenge of AI is far from technical capability, but close to cognitive governance. So... If AI is becoming part of our cognitive process, then AI literacy becomes a core capability, not a nice-to-have. People need to understand: • how AI systems reason • where they fail • when to trust them • and when to challenge them Links: - “Who’s in Charge? Disempowerment Patterns in Real-World LLM Usage.” https://lnkd.in/emi-DVQp - “Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender.” https://lnkd.in/eVPb3pQv - Thoughtful analysis published by My Tech Plan: https://lnkd.in/eJuSxmV5 #GovernYourAINow #GobiernaYaTuIA

Trust determines how AI is actually used. This is a lesson about human-system interaction that human factors psychology taught us long ago. Too many AI discussions focus mostly on accuracy. If a system performs well, people assume it will be used well. That is not what happens in practice. People do not respond to AI based on capability alone. They respond based on trust. When trust is too low, systems are underused. People ignore outputs or redo the work. When trust is too high, systems are over-relied on. Outputs are accepted without enough scrutiny. Errors pass through. Neither outcome is what we want. The goal is not maximum trust, nor is it minimal trust. It is calibrated trust. Trust is psychological. Trust is shaped by more than accuracy. Trust is shaped by how the system is introduced, how transparent it is, and how costly errors are (including how embarrassing they may be). Trust is also shaped by how the work is structured around it. Poorly designed workflows create constant monitoring and correction. Trust erodes. Clear roles and decision points stabilize use. Trust holds. The same system can be trusted in one setting and resisted in another. If we want effective use of AI, we have to design for trust. That is a psychological problem as much as a technical one. #appliedpsychology #psychology #humanfactors #artificialintelligence Brandon May Ph.D Emanuel Robinson Fred Oswald Mindy Shoss David Blustein

🏄 Human-AI Alignment Is About Trust Calibration AI decision making is a very sensible topic. One key question is, what makes AI assistance actually useful in human decision-making? The paper, “Learning to Decide with AI Assistance under Human-Alignment” (Benz et.al, 2026) explains that effective human-AI collaboration depends not only on model accuracy, but on alignment between human and AI confidence structures. The paper shows that humans do not automatically benefit from AI confidence scores. AI assistance becomes effective only when AI confidence meaningfully corresponds to how humans assess and interpret uncertainty within a decision context. Using both theoretical analysis and human-subject experiments, the authors demonstrate that aligned confidence structures significantly improve the ability of humans to learn when: - AI predictions should be trusted, - human judgment should dominate, - and additional verification becomes necessary. A key contribution of the paper is the shift from evaluating isolated model performance, to evaluating relational decision dynamics between humans and AI systems. In all domains where humans remain operationally responsible for decisions influenced by probabilistic AI recommendations this is an important observation. The paper frames human-AI alignment as a practical governance and system-design challenge involving trust calibration, uncertainty communication, cognitive interpretation and in particular workflow integration. 🎯 Bottom line: The paper shows that effective human–AI decision-making depends not only on AI accuracy, but on alignment between AI confidence and human confidence, making trust calibration a central design and governance challenge. 🔗 to the paper in the comments #Governance #HumanAI #ArtificialIntelligence #LegalTech #Trust

AI systems are becoming more autonomous and powerful, yet they still struggle to connect with the temporality that shapes human thought. Our cognitive rhythms, pacing, and timing patterns influence far more of our decision making than we usually realize, and when AI cannot align with these rhythms, collaboration often feels out of sync. This research grew from my curiosity about that very gap. Drawing on Allen’s temporal interval algebra and dual-process theories of cognition, I explore three fundamental forms of temporal reasoning that shape how humans interact with AI. These include how we understand intervals and durations, how we anticipate when events should occur, and how we interpret causal or sequential links between events. At the core of the study is the development of a real-time machine learning framework designed to identify a user’s temporal cognitive style from natural interaction data. When AI can adapt its timing to the individual, the collaboration becomes more intuitive. Interactions feel smoother, decisions align more naturally, and the system begins to support us in ways that respect how we process information over time. I believe this direction can lead to stronger trust, better coordination, and significantly improved outcomes compared to non-adaptive systems that overlook temporal cognitive diversity. For me, this work represents a step toward AI that not only understands what we think, but also how we think in time, making future human AI partnerships far more meaningful and effective. #AgenticAI #AIAgent #AI #Cognition

Trust comes from knowing AI’s limits. One of the most consistent surprises was how limited many lawyers’ understanding was of where generative AI struggles in practice, particularly with hallucinations and numeric reasoning. While participants knew outputs needed to be checked, few had a clear sense of when models were most likely to fail or how to interrogate those weaknesses effectively. Confidence increased when limits were made explicit, and lawyers were given practical ways to verify outputs. Instead of treating AI as something to either trust or avoid, participants learned how to supervise it. They learned to cross-check assumptions, review outputs critically, and build verification into their workflows. This resulted in more disciplined use rather than looser risk tolerance. Lawyers remained accountable, and trust became a function of judgment.

Do business teams rely on AI more than they should?⁣ ⁣ In most companies, AI is built by technical teams. It is then used by non-technical teams.⁣ That gap is where over-trust begins.⁣ ⁣ A dashboard shows “Risk Score: 94% Confidence.”⁣ ⁣ To an engineer, that means probability under certain assumptions.⁣ To a business user, it often feels like certainty.⁣ ⁣ And that difference matters.⁣ ⁣ >> 𝐓𝐡𝐞 𝐈𝐥𝐥𝐮𝐬𝐢𝐨𝐧 𝐨𝐟 𝐂𝐞𝐫𝐭𝐚𝐢𝐧𝐭𝐲⁣ Non-technical teams see:⁣ → Clean percentages⁣ → Green and red risk bands⁣ → Ranked lists⁣ → Clear recommendations⁣ ⁣ It feels definitive.⁣ ⁣ But a 94% model score does not mean 94% truth.⁣ It means the model is 94% confident based on past data.⁣ ⁣ → 𝐈𝐧 𝐜𝐫𝐞𝐝𝐢𝐭 𝐫𝐢𝐬𝐤, that 6% uncertainty may represent thousands of edge cases.⁣ ⁣ → 𝐈𝐧 𝐟𝐫𝐚𝐮𝐝 𝐝𝐞𝐭𝐞𝐜𝐭𝐢𝐨𝐧, even a 1% false negative rate can mean millions in losses at scale.⁣ ⁣ >> 𝐀𝐮𝐭𝐨𝐦𝐚𝐭𝐢𝐨𝐧 𝐁𝐢𝐚𝐬⁣ Automation bias is a documented cognitive effect.⁣ When a system suggests an answer, humans tend to follow it — especially under time pressure.⁣ ⁣ → 𝐈𝐧 𝐡𝐞𝐚𝐥𝐭𝐡𝐜𝐚𝐫𝐞, studies have shown clinicians are more likely to accept incorrect AI suggestions if they appear confident.⁣ ⁣ → 𝐈𝐧 𝐟𝐢𝐧𝐚𝐧𝐜𝐞, analysts reviewing flagged transactions often default to the model’s recommendation when volumes are high.⁣ ⁣ The pattern is simple:⁣ - AI suggests⁣ - Human confirms⁣ - Accountability blurs⁣ ⁣ The model becomes the decision-maker.⁣ The human becomes a rubber stamp.⁣ ⁣ >> 𝐓𝐡𝐞 𝐁𝐥𝐚𝐜𝐤 𝐁𝐨𝐱 𝐄𝐟𝐟𝐞𝐜𝐭⁣ Most non-technical teams do not see:⁣ → Training data limitations⁣ → Feature engineering logic⁣ → Drift monitoring⁣ → Threshold calibration⁣ ⁣ They see output.⁣ ⁣ 𝐅𝐨𝐫 𝐞𝐱𝐚𝐦𝐩𝐥𝐞:⁣ A fraud team sees “High Risk.”⁣ They do not see that the model was trained on last year’s transaction patterns.⁣ ⁣ If user behavior shifts, the model may still look confident while being wrong.⁣ Confidence without transparency builds blind trust.⁣ ⁣ 𝐌𝐲 𝐓𝐡𝐨𝐮𝐠𝐡𝐭𝐬⁣ AI should support judgment, not replace it.⁣ Because when non-technical teams over-trust AI, accountability does not disappear. It shifts back to leadership.⁣ ⁣ Do your business teams understand the limits of the AI systems they rely on? Share below.⁣ ⁣ To avoid over-trusting AI, always look for the trust scores. I have covered it here.⁣ ⁣ https://lnkd.in/dQfiyN7S ⁣ ⁣ 🔁 Repost if this helped you.⁣ ⁣ #EnterpriseAI #AIRisk #ResponsibleAI #AIGovernance #AIAdoption