HAAM doctrine / Summer Season '26
Human limits
Attention, memory, confidence, time, data, institutions, and the living world are finite. Good design does not pretend otherwise. It decides what the system should absorb, what evidence people must see, and where human judgment must remain.
The starting point
A product is always used under real conditions: someone is distracted, tired, unfamiliar with the subject, moving through a slow connection, using assistive technology, worried about consequences, or simply short on time. These are not unusual users. They are people.
Human limits are not defects to fix. They are conditions a responsible system must respect.
Three fields of constraint
Human capacity is only one part of the picture. Systems are uncertain and organizations are finite. Products also consume energy, materials, labor, maintenance, and ecological capacity.
01 / Human limits
Attention, working memory, perception, confidence, expertise, sensory and motor ability, time, and energy all vary by moment and person.
Sequence demands, preserve context, use plain language, support assistive technology, protect important decisions, and make recovery obvious.
02 / System limits
Data is incomplete. Models are uncertain. Permissions, latency, ownership, maintenance, organizational capacity, and accountability shape what a system can safely do.
Show provenance and uncertainty, define boundaries, design failure paths, preserve human handoff, and never hide responsibility behind automation.
03 / Living limits
Software depends on energy, minerals, devices, logistics, human labor, maintenance, ecosystems, and consequences that may appear far from the interface.
Treat extraction, consumption, longevity, repair, and ecological cost as design constraints rather than invisible externalities.
What design decides
The goal is not to remove the human. It is to remove avoidable cognitive and operational burden while keeping judgment, evidence, permission, and accountability in the right places.
01
What context should the product preserve instead of asking the person to reconstruct it?
Keep state, history, progress, and the next step visible.
02
What does someone need to know before they can act with confidence?
Explain enough for the decision, then reveal depth progressively.
03
What genuinely deserves scarce attention right now?
Protect important signals and quiet everything that can wait.
04
Where is human judgment necessary because values, risk, or tradeoffs remain?
Do not compress consequential choices into an opaque default.
05
What can be automated safely, and inside which explicit boundaries?
Move from observing to recommending to acting only as permission grows.
06
Can the person inspect evidence, interests, uncertainty, and why the system intervened?
Keep provenance, independence, and contestability visible.
07
What happens after misunderstanding, failure, or a wrong model output?
Provide stop, correct, undo, retry, and human escalation paths.
Proportional friction
Remove friction from reversible, low-risk work. Add a deliberate pause where consequences are high, evidence is weak, or permission is changing. The aim is not a frictionless product. It is a product whose friction matches the decision.
Research lineage
It emerged through reading, participatory design, fieldwork, a six-year research project, and the practical work of keeping products understandable and working.
Coursework on human-algorithm interaction connected transparency, fairness, accountability, validity, uncertainty, and trust. The underlying model belongs to the cited researchers; the durable lesson was that model output is only one part of the experience.
Participatory and service-design work in Tainan foregrounded situated context, different levels of technological confidence, mutual learning, autonomy, clear information, and the right of affected people to shape decisions.
A survey retaining 675 Taiwanese student responses and 30 individual prototype sessions showed that discoverability, trust, timing, information density, and action labels could matter more than generative capability. Useful information still fails when people do not notice it, understand it, or trust why it appears.
Attention design, interaction contracts, accessible product states, governed automation, evidence trails, and designed failure paths now apply the same principle across products and organizations.
Use the doctrine
A human-limits audit turns the philosophy into product decisions. It asks what the person is being required to carry, what the system can absorb, and what must remain visible.
Follow the argument
Design for actual humans
HAAM can map the demands, remove avoidable burden, preserve the decisions that need a person, and make the limits of the technology visible before they become failures.
Start a project ↗Optional analytics. Google Analytics and Clarity load only if allowed; form, email, and chat content are excluded.