Cognitive Debt vs Cognitive Stamina with AI If the machines don't rise up against us or simply displace us from our jobs, leaving us with a utopia of free time, then this is likely the next great generation correction we will be required to face. Today we are adjusting our collective beliefs on the smartphone and social media with emerging guardrails and policy on the age at which we access these mediums. In the near future that debate is certain to expand to how AI is impacting the cognitive development of our children and our own aging mental health. There is a recent and well publicised MIT study that has begun the work to observe the effects of using AI - in this case essay writing. While this isn't every use case for AI, its a fair and representative task requiring critical thinking, creativity and the ability to communicate. The study found: - **Reduced Neural Engagement**: LLM users showed significantly weaker alpha and beta band connectivity- brain rhythms associated with semantic processing and executive control compared to both Search and Brain-only groups [MIT Study, p. 78–80]. - **Memory Encoding & Ownership**: LLM users struggled to quote from their essays and reported lower ownership, even though their essays scored well by AI judges [MIT Study, p. 137–138]. - **Cognitive Offloading Effects**: When LLM group participants were asked to write without AI, their neural engagement did not return to Brain-only group levels. Their cognitive systems had adapted to the tool, and some capacities appeared diminished. [MIT Study, p. 106–108]. The study highlights the fine balance between cognitive modes: externally scaffolded automation versus internally managed curation. It is evident from the research performed that **repeated use of AI tools can create 'cognitive debt'** - the deferring of mental effort with observed short and potentially long-term costs. Before I got to the paper itself - the debate featured on Cal Newport's *Deep Questions*. (Cal being the advocate of 'Deep Work' and an influence on my own personal systems). Cal Newport likens this shift to the industrial revolution: as machines replaced physical labor, we invented gyms. *If we outsource mental labor to AI, will we need cognitive gyms just to stay sharp?* Cal extends the idea to reflect on how we've adapted to remain physically fit by creating a whole industry around personal fitness, gym memberships and the like. Should we give up our mental exercise in the workplace, then we might need to find the sort of 'cognitive gyms' that keep our brain muscle strong. While undoubtedly, the capitalistic nature of our world economy will demand these tools take up in our workplace (as the forklift once did) - we can make a choice about this in our arenas of learning, in schools and at home. But I don't think its all downside. So I'll offer some balance to the equation. Not to dispute any of the argument warning of cognitive debt - I think this is the right framing for a problem we need to address - but as a small token of hope and an example of personal experience where we can take advantage of this technology. **Cognitive Stamina** is the mental capacity to sustain focused, effortful thinking over extended periods of time without significant decline in performance, attention, or motivation. We all suffer this at some point in the working day - more it seems, earlier and earlier due to the fast paced, context switching, remote world we have created post-pandemic. While a reform in working expectations and an emphasis on developing personal cognitive stamina organically in the workplace would be preferred - AI provides the augmentation to support us and the economic model which thrives on productivity. I see 3 distinct benefits of AI: 1. **Creativity** - I have long been outspoken about my beliefs in the machine as a useful partner, and companion and collaborator in ideation. Not as the source of new ideas but as that digital reflection that can rapidly build on my new threads and create connections I would otherwise be unsighted to. In this mode, I retain the critical thinking elements but reaching step after step in the development process is accelerated by the digital tools. The MIT paper even supports such a claim: Brain-to-LLM group showed spikes in connectivity when AI was introduced after unaided writing [MIT Study, p. 121] 2. **Delegation** - I've written before about the spectrum of writing/communication tasks versus the usefulness of AI. On that scale, I observe that the more you care or are deeply invested in a topic, then the harder it is for AI to do a good job in either assisting you or writing it for you. The same applies ethically, for messages you know would be perceived differently if the recipient knew it was more than less written automatically. The other phenomenon I have personally experienced is reduction in reliance upon AI the more I write - coming from a place of barely authoring anything at all, AI has been the training wheels/stabilisers in this metaphor for me. So like any good leader or team player you would appropriately distribute tasks to others in order to free up your own capacity by not engaging all the way on things that others can do. 3. **Productivity** - Here's the big and maybe more controversial one. Cal Newport, and others, frequently advocate for a weekly design system that puts aside undistracted time for deep work fist thing in the morning, when we are most engaged and able to focus on important work. I've adopted this approach to good affect, but it does greater highlight the cognitive fog and strain later in the day as meetings and last minute tasks turn up and requiring context switching and effort when the reserves are low. AI could be a solution to this. For many tasks, following the 'rules' above in 1 and 2, AI can take over to extend and augmenting your capacity, giving you the digital energy for the last mile. I have found that completing those lower value, high-drain tasks at the end of the day with AI-assistance is worth more than pushing them to the next day where they compete and potentially displace time set aside for deep work. I haven't the resource to perform the study in the same way, to know what's going on in the brain at a biological level. Furthermore, the advantage I am observing may be limited to a generational segment of us who have had the chance to grow a decent cognitive muscle to begin with - before amplifying it with AI. Cognitive offloading can either create dependency or capacity, but I believe that alongside clear safeguarding of it's use, it is possible to design sustainable and additive working patterns that help us create space for greater organic cognitive work. *Where in your workflow are you building cognitive stamina vs borrowing it?* --- ### References 1. 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_. MIT Media Lab. Published June 2025. https://arxiv.org/abs/2506.08872 2. Newport, C. (2025). _Deep Questions with Cal Newport – Episode #359: Should we fear cognitive debt? Retrieved from www.thedeeplife.com/podcasts/episodes/ep-359-should-we-fear-cognitive-debt