Ethical AI Scheduling: Designing for Fluctuating Energy Levels

Most productivity tools are built on a fiction: that you are the same person every day. That the task you scheduled Tuesday evening — with clarity, with intention — will be equally within reach Tuesday morning. For millions of neurodivergent adults navigating ADHD, Autism, AuDHD, or acquired executive dysfunction, this fiction is not just unhelpful. It is exhausting.

Ethical AI scheduling begins with a different premise: energy is not static, and neither is cognitive capacity. Designing for fluctuating energy levels is not an accommodation. It is the baseline standard for any AI system that claims to support human wellbeing.

Why Traditional Schedulers Fail People with Executive Dysfunction

The conventional digital calendar has changed remarkably little since the paper planner it replaced. It presents a flat grid. It treats each time slot as identical. It assumes that marking something as "important" is sufficient motivation to begin it.

For individuals with executive dysfunction — a core feature of ADHD, Autism, and dozens of neurological and psychiatric conditions — this model fails at the most fundamental level. The barrier is not knowledge of what to do. It is the initiation gap: the neurological distance between intending to act and being able to begin. No reminder notification closes that gap. No priority flag resolves it.

When an AI scheduling system applies the same flat-list logic at scale — adding a machine's speed to an already broken model — it does not solve the problem. It amplifies it. More suggestions surfaced at the wrong moment, in the wrong tone, with no awareness of a person's current nervous system state, is not smarter technology. It is the same failure, faster.

The Allostatic Load Problem: A Scientific Framework for Fluctuating Capacity

The science of why capacity fluctuates is well-established. Allostatic load — a term from stress physiology — refers to the cumulative biological cost of chronic stress adaptation. When allostatic load is high, the nervous system is operating at or near its regulatory ceiling. Executive function — planning, initiation, cognitive flexibility — is precisely the set of capacities that degrades first.

Many neurodivergent adults describe allostatic load that accumulates faster and recovers more slowly than they observe in neurotypical peers — a clinical picture consistent with research on autistic burnout and chronic sensory load (Raymaker et al., 2020). Extending McEwen's allostatic load framework (McEwen, 1998; McEwen & Stellar, 1993), we model this asymmetric accumulation as a primary design constraint, even where direct comparative measurement in neurodivergent populations remains preliminary. A single morning of masking in a neurotypical environment can consume the executive capacity that, for another person, might be spread across an entire day. The calendar does not see this. Standard productivity AI does not see this either.

Ethical AI scheduling must model allostatic load as a first-order input — not a footnote, not an accessibility option buried in settings, but the primary signal that governs every suggestion the system generates.

The LALI Engine: Ethical AI Scheduling Built on Real-Time State

HolosCognitive's core suggestion system, the LALI engine (Logixr Allostatic Load Index), was designed around exactly this requirement. Rather than reading a task list and surfacing the next item, the LALI engine reads the user's current state before generating any suggestion at all.

The inputs to the LALI engine include user-reported somatic states — categorized as Prismatic, Fragmented, or Shards, each corresponding to a level of nervous system regulation — alongside a Capacity Index, a 0-to-1 float derived from somatic state history and check-in data. The allostatic load score is then computed from somatic state trajectories over configurable trailing time windows. Only after reading this full picture does the engine produce ranked suggestion sets calibrated to what the user is actually capable of right now.

This is what distinguishes ethical AI scheduling from conventional AI productivity tools. The LALI engine does not ask, "What is on the list?" It asks, "What is this person able to engage with in this moment?"

And critically: the LALI engine is a suggestion system, never an automation engine. It does not execute actions. It does not automate decisions. It does not override user agency. Every output is an option, not an instruction. When allostatic load is low and capacity is high, the system can surface a fuller range of tasks. When load is high and capacity is constrained, suggestion density is automatically reduced. The human retains full decision authority at every point.

Designing for the Limits: The Governor and Sanctuary Mode

What an AI system chooses not to do is as ethically significant as what it does. This principle is operationalized in HolosCognitive through a constraint layer called the Governor.

When a user enters Sanctuary Mode — a state corresponding to extreme allostatic load — the Governor suspends all LALI task suggestions entirely. No productivity nudges. No gentle reminders. Only co-regulation and grounding prompts are surfaced. The system recognizes that the most ethical action, in that moment, is to step back.

When somatic state registers at the Shards level, the Governor limits task output to a single, lowest-friction suggestion. Not a prioritized list. Not three options. One. The cognitive cost of choosing between options is itself a form of demand, and ethical AI scheduling accounts for even that overhead.

In Rupture State, all task-related outputs are suppressed entirely. The user is routed to regulation resources. The system carries a model of when it should stop functioning as a productivity tool — and it uses that model without requiring the user to ask.

A scheduling platform that does not know how to be quiet has not solved the problem of executive dysfunction. It has repackaged it.

The PDA Principle: Autonomy as a Non-Negotiable Design Standard

For users with a Pathological Demand Avoidance (PDA) profile, the design requirements of ethical AI scheduling become even more specific. Demand avoidance in the PDA context is a neurological drive to resist perceived demands, regardless of their source — including well-intentioned AI suggestions.

HolosCognitive's interface architecture responds to this directly by eliminating directive language, removing countdown timers and urgency cues, and presenting all LALI outputs as low-pressure options that can be dismissed without consequence. There are no streaks to maintain. No penalties for inaction. No leaderboards measuring productivity against other users. The platform does not gamify executive function. It scaffolds it.

This is not a feature. It is an ethical baseline for any platform claiming to serve neurodivergent users. Gamification mechanics — designed to leverage anxiety and loss-aversion to drive engagement in general user populations (Hamari, Koivisto, & Sarsa, 2014) — are not neutral design choices when applied to populations for whom anxiety is already a primary barrier to function. For users with a PDA profile in particular, externally imposed point systems and streak penalties read as demands (Newson, Le Maréchal, & David, 2003; O'Nions, Christie, Gould, Viding, & Happé, 2014), which is exactly what the platform is designed to avoid.

What Ethical AI Scheduling Demands of the Field

We are at a critical inflection point in the design of AI-powered scheduling and planning tools. The pattern currently emerging — "smarter" task managers that apply AI suggestion layers on top of conventional productivity logic — is not meaningful progress for the populations who need it most. It is the same neurotypical framework with a new surface.

Ethical AI scheduling requires us to hold several non-negotiable premises simultaneously: that cognitive capacity is dynamic and must be measured in real time; that the act of deciding what to do next can itself be the primary barrier; that an AI system has a responsibility to know when silence is the correct output; and that autonomy is not a preference setting — it is a design requirement.

HolosCognitive is built from these premises, not toward them. The distinction matters. A platform designed from the ground up to honor fluctuating executive function produces a fundamentally different architecture than one retrofitting accessibility features onto a neurotypical core. The allostatic load model, the Governor, Sanctuary Mode, the single-task Shards limit — these are not edge-case accommodations. They are the center of the design.

We need AI systems that hold allostatic load as a first-class input. That understand the ethics of suggestion, not just the mechanics. That treat the decision to stay quiet as a feature, not a failure. That is what designing for fluctuating energy levels actually demands — and it is the standard we should hold the entire field to.

---

References

• Barkley, R. A. (1997). Behavioral inhibition, sustained attention, and executive functions: Constructing a unifying theory of ADHD. Psychological Bulletin, 121(1), 65–94. https://doi.org/10.1037/0033-2909.121.1.65
• Hamari, J., Koivisto, J., & Sarsa, H. (2014). Does gamification work? — A literature review of empirical studies on gamification. In Proceedings of the 47th Hawaii International Conference on System Sciences (pp. 3025–3034). IEEE.
• McEwen, B. S. (1998). Stress, adaptation, and disease: Allostasis and allostatic load. Annals of the New York Academy of Sciences, 840(1), 33–44. https://doi.org/10.1111/j.1749-6632.1998.tb09546.x
• McEwen, B. S., & Stellar, E. (1993). Stress and the individual: Mechanisms leading to disease. Archives of Internal Medicine, 153(18), 2093–2101.
• Newson, E., Le Maréchal, K., & David, C. (2003). Pathological demand avoidance syndrome: A necessary distinction within the pervasive developmental disorders. Archives of Disease in Childhood, 88(7), 595–600. https://doi.org/10.1136/adc.88.7.595
• O'Nions, E., Christie, P., Gould, J., Viding, E., & Happé, F. (2014). Development of the 'Extreme Demand Avoidance Questionnaire' (EDA-Q): Preliminary observations on a trait measure for Pathological Demand Avoidance. Journal of Child Psychology and Psychiatry, 55(7), 758–768.
• Raymaker, D. M., Teo, A. R., Steckler, N. A., Lentz, B., Scharer, M., Delos Santos, A., Kapp, S. K., Hunter, M., Joyce, A., & Nicolaidis, C. (2020). "Having all of your internal resources exhausted beyond measure and being left with no clean-up crew": Defining autistic burnout. Autism in Adulthood, 2(2), 132–143. https://doi.org/10.1089/aut.2019.0079
• Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285. https://doi.org/10.1207/s15516709cog1202_4
• Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes (M. Cole, V. John-Steiner, S. Scribner, & E. Souberman, Eds.). Harvard University Press.