Exploring Psychosocial Safety in the Context of Workplace Health & Safety: What Leaders need to know

In most organizations, two conversations are happening in two different rooms.

In one, HR and People & Culture are talking about psychological safety: whether people feel able to speak up, admit mistakes, challenge ideas and ask for help. In the other, health and safety professionals are dedicated to identifying hazards. Some of those are psychosocial hazards: the parts of work design, management, and relationships that can cause psychological or physical harm. More and more, employers are being tasked with identifying and controlling psychosocial hazards to mitigate operational and legal risks. Safe Work Australia. (2022). Model Code of Practice: Managing psychosocial hazards at work. Safe Work Australia.

Yet, who is responsible? The owners, budgets, and reporting lines usually aren’t clear. And that gap is where risk hides.

A neuroscience-based approach to psychosocial risk management lets HR and Health and Safety Professionals work from the same playbook. This is what ABL has to offer.

Understanding the Context

  • Psychological safety is a team climate. It’s the shared belief that you won’t be punished or humiliated for speaking up with ideas, questions, concerns or mistakes. It’s a culture and performance concept.
  • Psychosocial safety is a risk-management obligation. It’s about identifying psychosocial hazards (such as high job demands, poor support, bullying or poor organizational justice), assessing the risk, and putting controls in place so far as is reasonably practicable, as you would for any other workplace hazard.
  • They’re connected through the brain. Both describe how people respond to social threats. When the brain detects a threat to things like certainty, control, fairness, status or belonging, it shifts into self-protection. That shift is what makes people go quiet (low psychological safety), and, sustained over time, it’s what turns a psychosocial hazard into psychological harm.

 

What is psychological safety?

“Teams with psychological safety share a belief… that they will not be punished or humiliated for speaking up with ideas, questions, concerns or mistakes, and that the team is safe for interpersonal risk taking” – Amy Edmondson, PhD.

From a neuroscience perspective, Psychological Safety is the state where the brain’s need for security, autonomy, fairness, esteem and trust are met in a social environment.

It matters because it’s one of the strongest predictors of team learning and performance. Google’s Project Aristotle, which studied hundreds of its own teams, found psychological safety was the most important factor separating its highest-performing teams from the rest.

But psychological safety is often misread as “being nice” or “keeping everyone comfortable”. It isn’t. Psychologically safe teams are frequently more challenging places to work, because candour is expected. What’s removed is the fear of personal consequences for being candid.

What is psychosocial safety?

Psychosocial safety comes from the world of work health and safety. In Australia, Health and Safety professionals laws require persons conducting a business or undertaking (PCBUs) to manage psychosocial risks the same way they manage physical risks: identify hazards, assess risk, control it, and review controls. Safe Work Australia’s Model Code of Practice: Managing psychosocial hazards at work sets out the common hazards, including:

  • Job demands that are too high or too low
  • Low job control
  • Poor support from supervisors or colleagues
  • Lack of role clarity
  • Poor organizational change management
  • Inadequate reward and recognition
  • Poor organizational justice
  • Traumatic events or material
  • Remote or isolated work
  • Poor physical environment
  • Violence and aggression
  • Bullying
  • Harassment, including sexual and gender-based harassment
  • Conflict or poor workplace relationships

Internationally, ISO 45003 gives similar guidance for managing psychological health and safety within an occupational health and safety management system.

The key point for leaders: psychosocial safety is not optional, and it’s not just a wellbeing initiative. It’s a duty, and regulators increasingly expect to see evidence of how hazards are being identified and controlled.

Why the difference matters

When organizations treat these as separate topics, three things tend to happen.

  1. Compliance without culture. Health and Safety professionals teams build a psychosocial risk register, run a survey and update policies. All necessary. But if people don’t feel safe to tell the truth, the survey under-reports, consultation is polite rather than honest, and hazards stay hidden until they surface as a complaint, a resignation or a claim.
  2. Culture without controls. HR runs psychological safety workshops and wellbeing programs. Also valuable. But if workloads, rosters, role clarity and change processes stay the same, the hazards remain. Training people to feel safe in an unsafe system doesn’t hold.
  3. Nobody owns the gap. When a psychological injury claim lands, HR and Health and Safety professionals often discover they were each seeing part of the picture. Neither had the whole story early enough to act.

The organizations that get this right treat psychological safety as the early-warning system for psychosocial risk, and psychosocial risk management as the structure that turns what they hear into action.

The missing link: how the brain processes social threat

Neuroscience explains why these two ideas belong together.

The human brain is constantly, and mostly nonconsciously, scanning for threat and reward. It doesn’t only respond to physical danger. It responds just as strongly to social signals: a public put-down, an unfair decision, a sudden restructure, being left out of a meeting. Research suggests social pain and physical pain share overlapping neural pathways, which is why a harsh comment can feel like a blow.

When the brain detects social threat, it does what it’s built to do. It narrows attention, prioritises self-protection, and reduces the risky behaviour of speaking up. Short term, that’s psychological safety dropping. Long term, chronic threat keeps the stress response switched on, which is linked to burnout, anxiety, poorer physical health, more time off work and, eventually, employee turnover.

Put simply: a psychosocial hazard is anything at work that keeps a person’s brain in sustained threat. Psychological safety is what it feels like when those threats are low enough for people to speak, learn and perform.

The S.A.F.E.T.Y.™ model: one framework for both rooms

The Academy of Brain-based Leadership (ABL) developed the S.A.F.E.T.Y.™ model, drawing on two decades of neuroscience research, to make these social threats visible and measurable. It identifies six domains the brain is continually tracking:

The S.A.F.E.T.Y.™ domains and what the brain is tracking and where it shows up as a hazard
Domain What the brain is tracking Where it shows up as a hazard
Security Predictability and certainty Poor change management, lack of role clarity and unclear expectations
Autonomy A sense of choice and control Low job control and micromanagement
Fairness Equitable and just treatment and transparency Inconsistent process, poor organizational justice, lack of opportunity to contribute, and unfair decisions
Esteem Status, respect and recognition Inadequate recognition, feeling undervalued or judged
Trust Social need to belong Isolation and exclusion
You Unique personal context One-size-fits-all approaches that miss individual triggers

Because the same drivers sit underneath both team climate and psychosocial hazards, the model gives HR and Health and Safety professionals a shared language. A Health and Safety professionals manager looking at “poor organizational justice” and an HR leader worried about “people not speaking up” may be looking at the same Fairness threat from two angles.

What this looks like in practice

For HR and People & Culture

  • Move from measuring sentiment to measuring the drivers of threat and reward.
  • Use individual S.A.F.E.T.Y.™ profiles so leaders understand that what threatens one person (say, loss of Autonomy) may barely register for another, who is more sensitive to Fairness.
  • Build leader capability in small, repeatable behaviours and specific data relating to their team, rather than one-off workshops.

For Compliance Officers, Health and Safety professionals, and Occupational Health

  • Treat low psychological safety as a leading indicator: if teams won’t speak up, your hazard data is incomplete.
  • Use S.A.F.E.T.Y.™ diagnostics, like the team level PS Pulse™, alongside surveys and incident data to find hotspots earlier.
  • The PS Pulse™ is an anonymous diagnostic that pinpoints the human S.A.F.E.T.Y.™ driver being threatened in a team, so controls address the cause, not just the symptom. Utilize the diagnostic over time to review changes and improvements, to ensure controls are current.

Together

  • Agree one shared set of team-level measures.
  • Map hazards on the risk register to S.A.F.E.T.Y.™ domains.
  • Review progress together, with evidence you can show leadership, the board and, if needed, the regulator.

How ABL supports HR and occupational health teams

ABL’s corporate offerings are designed to fit this diagnose → intervene → monitor cycle:

  • S.A.F.E.T.Y.™ Assessment — a validated self-diagnostic (about 15 minutes) that shows which social drivers matter most to each person and how they experience threat and reward.
  • PS Pulse™ — an anonymous team diagnostic that surfaces where psychological safety is under strain and which drivers are involved.
  • Individual and team S.A.F.E.T.Y.™ debriefs and toolkit — practical sessions that build self-awareness, interpersonal awareness and leader micro-skills.
  • ABL SAFETY™ Dashboard — ongoing visibility of team-level trends.
  • The S.A.F.E.T.Y.™ System — a framework for embedding the approach across the organization.

ABL works with more than 100 organizations across healthcare, pharmaceuticals, financial services, tech, energy, manufacturing, education, entertainment, fashion and apparel, professional services, food and beverage, consumer staples, transportation, and not for profit sectors.

AI Can’t Fix a Psychologically Unsafe Culture at Work

You’ve probably seen the headlines. AI coaching tools promise to revolutionize leadership development. They offer 24/7 availability, instant feedback, and data-driven insights at a fraction of the cost of human coaches. Some platforms claim they can even detect emotional states through voice analysis or predict team dynamics through sentiment tracking.

Impressive? Absolutely.

A replacement for building genuine psychological safety? Not even close.

Here’s the uncomfortable truth: AI can analyze patterns, but it cannot create the conditions where people feel genuinely safe to be themselves. It can suggest responses to conflict, but it can’t model vulnerability. It can track engagement metrics, but it can’t build trust through shared human experience.

And if your team culture already lacks psychological safety, adding AI coaching tools might actually make things worse.

The Illusion of Connection

A hot take on AI coaching tools and the irreplaceable value of genuine psychological safety

Let me paint a picture. Sarah, a mid-level manager, receives a notification from her company’s new AI coaching platform. It’s detected that her team’s communication patterns suggest “potential disengagement.” The AI recommends she schedule one-on-ones and provides scripted questions to ask.

Sarah follows the script. Her team members give polite, surface-level answers. The AI marks the intervention as “complete.” The dashboard shows green checkmarks.

But nothing has actually changed.

Why? Because Sarah’s team doesn’t lack structured conversations. They lack the felt sense that it’s safe to tell Sarah what’s really going on. They need to know that admitting struggle won’t be held against them, that asking for help won’t signal incompetence, that disagreeing won’t damage their standing.

That kind of safety isn’t built through algorithms. It’s built through repeated human interactions where leaders demonstrate, through their own behavior, that vulnerability is not only accepted but valued.

What AI Misses: The Neuroscience of Safety

At ABL, our S.A.F.E.T.Y.™ framework is grounded in decades of neuroscience research. We know that psychological safety isn’t just a nice-to-have cultural feature. It’s a neurobiological necessity for high performance and innovation.

The framework identifies six core drivers that our brains need to function optimally:

  • Security: The need for certainty,
  • predictability, and consistency
  • Autonomy: The need for a sense of choice and control over our lives
  • Fairness: The need for equitable treatment and transparent processes
  • Esteem: The need to feel valued and respected
  • Trust: The need to belong to a trusted group and experience reliable relationships with those ‘like us
  • You: The unique factors that shape our individual experience

When these needs go unmet, our brains shift into threat mode. The amygdala takes over. Creativity shuts down. Collaboration becomes difficult. Performance suffers.

AI tools can measure some symptoms of these unmet needs such as declining engagement scores, increased turnover, reduced communication frequency– all lagging indicators of a problem when it’s too late. However, AI can’t address the root cause because it cannot provide the genuine human connection that the human brain fundamentally requires, nor can it predict, perceive, or contribute to a human’s response to their ‘human’ social experiences and interactions.

When you connect with a person, on a real level your brain might release oxytocin—the neurochemical associated with bonding and trust. Oxytocin is implicated in pair bonding, parental behavior, affiliation, and some forms of trust.

It’s quite different when an algorithm validates your feelings.

While an AI Chat Bot’s algorithmically delivered digital feedback or affirming responses can, indeed, engage reward-related brain regions, actual reciprocal human relationships are a particularly robust source of belonging and support that still cannot be matched by tech.

Where AI Falls Short on Each S.A.F.E.T.Y.™ Driver

Let’s break this down further.

Security Security requires more than risk assessments and compliance tracking. People feel secure when they experience predictable patterns, consistent processes, and clear information during uncertain times. When a team member asks about a change, they need detailed explanations, advance notice, and assurance that core systems will remain stable. AI can provide information, but it can’t create the felt sense of environmental stability and predictability that comes from a leader who maintains consistent routines, follows through on commitments, and minimizes disruptive surprises.

Autonomy isn’t just about being consulted—it’s about having genuine control and choice over your work and matters that impact you and your life. When a leader gives you decision-making authority over how you accomplish your goals, you feel your autonomy honored. AI might offer options, but it can’t grant the real authority and control that makes autonomy meaningful.

Fairness depends on transparency and the ability to explain the “why” behind decisions in ways that account for context and nuance. AI can apply rules consistently, but fairness often requires understanding the human story behind the situation. It requires judgment calls that balance competing values, something humans excel at and machines struggle with.

Esteem is built through recognition that feels personal and specific. Generic praise generated by an algorithm doesn’t land the same way as a leader who notices the particular way you solved a problem or the unique strength and value you brought to a project. Your brain knows the difference between authentic appreciation and automated praise and acknowledgment.

Trust develops through belonging to a group where we feel an ‘in-group’ connection —where we’re recognized as part of the tribe. AI can facilitate connections, but it can’t create the felt sense of tribal belonging that comes from being recognized as ‘one of us’ by other humans who share our identity, values, or experiences.

You—each person’s unique background, personality, culture, and triggers shape how they experience Security, Autonomy, Fairness, Esteem, and Trust. AI might personalize responses based on data, but it can’t truly understand the complex interplay of someone’s life story, cultural context, and personal history that determines what makes them feel psychologically safe.

The Real Risk of Human–Robot Interactions (HRI)

Human–human interactions (HHI) and human–robot interactions (HRI)

Here’s where things get concerning. Organizations struggling with psychological safety often look for quick fixes. AI coaching tools can seem like an efficient solution. They are scalable, measurable, and cost-effective. Yet deploying these tools in a psychologically unsafe culture can backfire in several ways, such as:

False sense of progress: Leadership sees adoption metrics and completion rates, assuming the problem is being addressed. Meanwhile, team members go through the motions without experiencing any real shift in how safe they feel.

Increased surveillance anxiety: AI tools that monitor communication patterns, sentiment, or productivity can heighten feelings of being watched and judged. For team members who already don’t feel safe, this amplifies their Security, Autonomy, Fairness, Esteem, and Trust concerns.

Depersonalization of support: When people reach out for help and receive algorithmic responses, it can reinforce the feeling that they’re just data points rather than valued individuals. This directly undermines their Esteem and  Trust.

Erosion of Trust: If AI tools make recommendations that leaders follow without genuine understanding or adaptation, it becomes obvious that the leader is outsourcing their judgment. This damages Trust in leadership.

Authenticity gap: Perhaps most damaging, AI tools can create a veneer of care without the substance. Team members can tell when responses are scripted or when check-ins are prompted by software rather than genuine interest or concern. This gap between appearance and reality corrodes the foundation of psychological safety.

What AI Can Do (And What Humans Must Do)

This isn’t an anti-technology argument. AI has legitimate value in leadership development and team management.

AI excels at:

  • Identifying patterns in large datasets that humans might miss
  • Providing just-in-time learning resources
  • Offering practice scenarios for skill development
  • Tracking progress on measurable behaviors
  • Reducing administrative burden so leaders have more time for people

These are useful capabilities. But they’re support tools, not replacements for human leadership.

Building psychological safety requires humans to:

  • Model vulnerability and authenticity
  • Respond to emotional nuance with empathy
  • Make judgment calls that balance competing values
  • Build trust through consistent, reliable presence
  • Create the felt sense of being seen and valued
  • Navigate complex interpersonal dynamics
  • Demonstrate that it’s safe to speak up, take risks and make mistakes

These capabilities emerge from our shared humanity. They require emotional intelligence, social awareness, and the willingness to be present with discomfort. Qualities that remain distinctly human.

A Better Approach: Human-Centered, Tech-Supported

The most effective path forward combines human wisdom with technological support.

Start with the human foundation. Before they can create safety for others, leaders need to understand their own S.A.F.E.T.Y.™ drivers, what their brain needs, how it is triggered, and how their S.A.F.E.T.Y.™ sensitivities may impact themselves and those they lead. This self-awareness comes through reflection, feedback from other humans, and the sometimes uncomfortable work of examining our own biases and blind spots.

Once leaders have built this foundation, they can use AI tools strategically:

  • To surface data that prompts important conversations
  • To provide reminders and structure for consistent practices
  • To offer learning resources that deepen understanding
  • To track whether behavioral changes are occurring over time

But the actual work of building psychological safety: the conversations, the vulnerability, the connection, the trust-building, remains human work.

The Path Forward for Leaders Like You

If you’re a team leader reading this, you might feel the pressure to adopt the latest tools while also sensing that something essential is missing from your team culture. You’re not wrong.

Your team doesn’t need more technology. They need more of you. The real, imperfect, trying-your-best you.

They need you to:

  • Ask how they’re really doing and wait for the honest answer
  • Admit when you don’t know something
  • Respond to mistakes with curiosity instead of judgment
  • Notice what’s unique and valuable about each person’s contributions
  • Follow through on commitments, especially small ones
  • Show up consistently, even when it’s uncomfortable

This is harder than implementing an AI tool. It requires ongoing effort, self-reflection, and the courage to be genuinely present with your team.

But this is also where the real transformation happens. When you create genuine psychological safety, you unlock your team’s full potential. People bring their best thinking, take smart risks, collaborate across differences, and stay engaged even when work gets challenging.

No algorithm can create that. Only humans can.

The Bottom Line

AI coaching tools have their place in the modern workplace. They can support learning, provide structure, and offer useful data. But they cannot, and should not, replace the human work of building psychological safety.

Your team’s brains are wired to seek connection with other humans. They need to feel seen, valued, and safe in ways that only genuine human interaction can provide. The S.A.F.E.T.Y.™ drivers aren’t just abstract concepts; they’re neurobiological requirements for optimal functioning.

Before you invest in the next AI solution, ask yourself: Have we built the human foundation? Do our leaders understand their own psychological safety needs? Can they recognize and respond to these needs in others? Are we creating the conditions where people feel genuinely safe to be themselves?

If the answer is no, no amount of artificial intelligence will fix that. But if you’re willing to do the human work. The messy, uncomfortable, deeply rewarding work of building real psychological safety. You’ll create something no algorithm can replicate: a team where people thrive.

Ready to Build Real Psychological Safety?

At ABL, we help leaders develop the self-awareness and skills needed to create genuinely psychologically safe teams. Our approach is grounded in neuroscience and focused on the human capabilities that make great leadership possible.

Explore our resources:

Because building psychological safety isn’t about finding the right tool. It’s about becoming the right kind of leader.

 

Citations

Burenkova OV, Dolgorukova TA, An I, Kustova TA, Podturkin AA, Shurdova EM, Talantseva OI, Zhukova MA, Grigorenko EL. Endogenous oxytocin and human social interactions: A systematic review and meta-analysis. Psychol Bull. 2023 Sep-Oct;149(9-10):549-579. doi: 10.1037/bul0000402. PMID: 38713749; PMCID: PMC11077008.

Carter CS. Oxytocin and love: Myths, metaphors and mysteries. Compr Psychoneuroendocrinol. 2021 Dec 27;9:100107. doi: 10.1016/j.cpnec.2021.100107. PMID: 35755926; PMCID: PMC9216351.

Delgado MR, Fareri DS, Chang LJ. Characterizing the mechanisms of social connection. Neuron. 2023 Dec 20;111(24):3911-3925. doi: 10.1016/j.neuron.2023.09.012. Epub 2023 Oct 6. PMID: 37804834; PMCID: PMC10842352.

De Castro Martins C, Chaminade T, Cavazza M. Causal Analysis of Activity in Social Brain Areas During Human-Agent Conversation. Front Neuroergon. 2022 May 17;3:843005. doi: 10.3389/fnrgo.2022.843005. PMID: 38235459; PMCID: PMC10790851.

Fang CM, Liu AR, Danry V, Lee E, Chan SWT, Pataranutaporn P, Maes P, Phang J, Lampe M, Ahmad L, Agarwal S. How AI and human behaviors shape psychosocial effects of chatbot use: a longitudinal randomized controlled study. arXiv [Preprint]. 2025 Mar 21. Available from: https://arxiv.org/abs/2503.17473

Spatola, N., Chaminade, T. Precuneus brain response changes differently during human–robot and human–human dyadic social interaction. Sci Rep 12, 14794 (2022). https://doi.org/10.1038/s41598-022-14207-9

AI and the Brain: Answering the Questions Everyone’s Asking

The uptake of AI has moved from a novelty to daily habit for hundreds of millions of people in the space of a few years. That speed has left the science trailing the adoption, but a wave of studies published over the past year are starting to give real answers to the questions people are actually asking. While the evidence in the neuroscience field is still in its infancy, there are some interesting themes emerging.  Below are some common questions I’ve encountered over the past year by concerned AI users from a variety of backgrounds and industries, with some associated early evidence.

Is AI making us worse thinkers?

Not exactly, but it does seem to be changing when and how hard we think. The clearest evidence comes from MIT Media Lab’s “Your Brain on ChatGPT” study, which had 54 participants write essays while wearing EEG headsets, either using ChatGPT, using a search engine, or working entirely unaided. The unaided group showed the strongest, most widespread neural connectivity during the task; the search-engine group was intermediate; the ChatGPT group showed the weakest overall neural coupling.  That doesn’t mean people using AI produced worse essays.  In fact, human and AI judges often scored the AI-assisted essays competitively. What changed was how much cognitive work went into producing them, and how well people retained what they’d made. In the first session, 83% of ChatGPT users couldn’t correctly quote a sentence from an essay they’d written minutes earlier, compared to about 11% in the other two groups.

If I stop using the AI, does my brain “snap back”?

Not immediately, according to the same study. A subset of participants was reassigned to the opposite condition in a fourth session — people who’d used ChatGPT for three sessions had to write unaided, and vice versa. The group moving from ChatGPT to unaided writing still showed weaker neural connectivity and reduced engagement, even with the tool taken away. The carryover suggests the effect isn’t purely situational — something about sustained AI reliance seems to persist for at least a little while after the tool is gone. The reverse was more encouraging: participants moving from unaided writing to ChatGPT showed a network-wide spike in connectivity and higher recall, suggesting the deficit isn’t necessarily permanent, just slow to reverse.

Do we actually trust AI too much as a society?

Most likely. A Wharton team ran three preregistered experiments with 1,372 participants, giving them reasoning problems where an AI assistant was sometimes deliberately programmed to produce wrong answers. When the AI was wrong, about 80% of participants followed it anyway, performing worse than if they’d had no AI assistance at all. Trust in the AI was the strongest predictor: people who trusted the tool more had roughly 3.5 times greater odds of following its faulty answers. Only around 20% of participants actively caught and overruled the AI’s mistakes. The researchers call this pattern “cognitive surrender”, which refers to a conscious decision to trust AI, but a quiet failure to scrutinize what it produces. A related survey of knowledge workers by Microsoft and Carnegie Mellon researchers found the same effect in real workplace use: higher confidence in generative AI tracked with less critical thinking, and a shift from active problem-solving toward passive review of AI output.

Does using AI to check facts make me better at spotting misinformation on my own?

Apparently not much. A follow-up MIT Media Lab study found that conversations with AI reduced people’s belief in misinformation in the short term but built no lasting ability to detect it independently. The researchers compare this to how GPS may have eroded our capacity to navigate without it….the tool solves the immediate problem without transferring the underlying skill.  So it may be that our brains are being wired to decrease the ability to objectively scrutinize the information around us.

Is AI making everyone think and write more alike?

There’s early evidence of this too. In the MIT essay study, ChatGPT-assisted essays converged on more similar language and content than essays from the unaided or search-engine groups.  In other words, participants using AI tended to reach for the same phrasing and the same named entities. A separate, broader paper in Trends in Cognitive Sciences looked at this at scale across large language model usage and found a similar homogenizing effect on human expression and thought. If AI systems are trained on similar data and tend toward similar “safe” outputs, it stands to reason that heavy reliance on them could narrow the range of how people express ideas, though this is an early and still-developing area of research.

Is this just a problem for students, or does it impact experts too?

Both, but in different ways. Most of the research above involves people still building skills, like students, general knowledge workers, developers. A separate study of computer program developers learning a new coding library with and without AI assistance and found that AI sped up task completion but measurably slowed how well developers actually learned the material. So, what we are seeing is an impact on the rate of skill formation.  The shortcut works, but the underlying competence doesn’t develop the same way.

A different and arguably more unsettling finding comes from a year-long study of practicing oncologists. AI assistance improved their diagnostic accuracy but appeared to affect their subjective confidence in their own clinical judgment. This isn’t a failure to learn, but something closer to “intuition rust” in people who already have the expertise. It suggests AI reliance can erode confidence, even in situations where it clearly makes people more accurate.

Does this mean I should stop using AI tools?

The research doesn’t point that way. One of the more useful findings comes from Michael Gerlich, a researcher whose earlier work is often cited as evidence that AI erodes critical thinking. In a follow-up study, he had 150 participants write essays under three conditions: no AI, unguided ChatGPT use, and ChatGPT use with structured prompting. His conclusion was that the offloading effect depends heavily on how people engage with AI, not simply whether they use it.  What it suggests is that deliberate, scaffolded use can support critical thinking rather than erode it. In simple terms, that means putting structure, guidance and human oversight around AI so that it improves people’s work without replacing their judgment. That distinction, passive offloading versus active, structured engagement is probably the single most actionable takeaway across this entire body of research.

How solid is all of this, really?

Considering how new all of this is, the data are reasonably solid in direction, but still early in certainty. The flagship MIT study is a preprint, not yet peer-reviewed, with a sample of 54 people (only 18 completed the crossover phase), testing one model (GPT-4o) on one task (short essay writing). The Wharton and Microsoft/CMU studies have larger samples and add real-world workplace data, but every study in this space shares the same limitation of being short-term. Sessions span weeks or months, not years.  That’s the biggest open question left in the field:  what happens over years of sustained AI use, rather than months, and how that plays out differently for adults with established cognitive habits versus younger users whose skills are still forming. Almost nothing exists yet on that timeline, and it’s the gap every author in this space points to next.

Across writing, reasoning, factchecking, coding, and medicine, a consistent pattern is emerging: AI tools that solve a problem for you tend to reduce the depth of engagement, memory formation, or confidence that comes from solving it yourself.  However, the size and duration of that effect look different depending on how deliberately the tool is used. None of the current research supports avoiding AI but rather suggests that you are likely doing your brain a disservice by putting total reliance and unconditional trust in its output.

 

References:

Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15(1), 6. https://www.mdpi.com/2075-4698/15/1/6

Gerlich, M. (2025). From Offloading to Engagement: An Experimental Study on Structured Prompting and Critical Reasoning with Generative AI. Data, 10(11), 172. https://doi.org/10.3390/data10110172

Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X. H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. arXiv preprint arXiv:2506.08872. https://arxiv.org/abs/2506.08872

Lee, H. P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ’25), Article 1121. https://doi.org/10.1145/3706598.3713778

Rani, A., Danry, V., Liang, P. P., Lippman, A., & Maes, P. (2026). Dialogues with AI Reduce Beliefs in Misinformation but Build No Lasting Discernment Skills. In Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI ’26), Article 792. https://doi.org/10.1145/3772318.3790656

Shaw, S. D., & Nave, G. (2026). Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender. Wharton School Research Paper / SSRN working paper. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6097646 (preprint: https://doi.org/10.31234/osf.io/yk25n_v1)

Shen, J. H., & Tamkin, A. (2026). How AI Impacts Skill Formation. arXiv preprint arXiv:2601.20245. https://arxiv.org/abs/2601.20245

Sourati, Z., Ziabari, A. S., & Dehghani, M. (2026). The Homogenizing Effect of Large Language Models on Human Expression and Thought. Trends in Cognitive Sciences. https://www.cell.com/trends/cognitive-sciences/abstract/S1364-6613(26)00003-3 (preprint: https://arxiv.org/abs/2508.01491)

Return to top