Horizon 07
Health & Human Systems
Where emerging capability meets human consequence, and where technical performance stops being the same thing as operational safety.
Where technical capability becomes human consequence.
The shift
Healthcare is where several of the questions this practice asks stop being abstract. A model that is wrong in a spreadsheet produces a bad number. A model that is wrong in a triage pathway produces a delayed diagnosis, and the delay is experienced by a person.
The same questions arise wherever human capability is extended, measured, or shared with a machine. A tool that performs well under evaluation can still worsen an outcome by changing what someone attends to, how quickly they act, or what they quietly stop checking themselves. That holds in a resuscitation bay, and it holds in a cockpit, an operations centre, or a control room.
This area is informed by direct clinical and operational experience — advanced-practice nursing and emergency medicine — rather than by observing the field from outside. High-consequence environments are unforgiving of the assumption that measured performance describes real behaviour. That mostly changes which questions seem worth asking.
A system does not need to fail completely to change an outcome.
What we ask
- What does this change about what the person in the loop attends to?
- Which independent check stopped happening once the system became trusted?
- How does this behave at three in the morning, mid-handover, on an incomplete record?
- If it were degraded rather than unavailable, how long before anyone noticed?
- When the recommendation is wrong and it is followed, where does responsibility sit?
Lines of inquiry
- 01
Judgment and decision support
Decision support does not simply inform a decision; it reframes it. We examine what happens to professional judgment when a recommendation arrives before the person has formed their own, and what it takes to evaluate a system that cannot be independently inspected.
- 02
The human–machine interface
Where a person and a system share control, the interface decides more than it appears to: when a recommendation is encountered, whether disagreement is practical, and who is understood to be operating rather than supervising. Alert burden, automation bias, and override rates determine whether a human-in-the-loop control is a safeguard or a formality — authority to override that is never exercised is not evidence of agreement. Mode confusion and unclear handover are interface problems long before they are human error.
- 03
Human performance and its measurement
Attention, fatigue, cognitive load, and skill are increasingly monitored, modelled, and optimised — in clinical work and well beyond it. Measurement changes behaviour. We are interested in what is actually being measured, what is being inferred from it, and what happens to the people on the other end of the inference.
- 04
Security as a patient-safety property
A healthcare intrusion is usually assessed as a confidentiality incident. Its operational effect is on medication delivery, diagnostics, imaging, communications, device availability, and patient flow. The useful question is whether care continues, not only whether records left the building.
- 05
From recommendation to action
As clinical software moves from suggesting toward initiating — ordering, titrating, escalating, scheduling — the questions become authority, reversibility, and who is able to stop it. Accountability models in this setting assume a clinician decided.
- 06
Assurance where performance is not enough
Regulatory clearance and validation performance describe a system under study conditions. Trust in practice also depends on monitoring, escalation, honest communication of uncertainty, oversight that is real rather than nominal, and controls that survive staff turnover.
- 07
The long life of health data
Health information stays sensitive for decades. That makes it one of the clearer cases where capture-now-decrypt-later is a rational adversary strategy rather than a theoretical one, and where cryptographic lifecycle becomes a governance question.
Applied Horizons conducts research and analysis. We do not provide medical advice, clinical care, or patient services. Nothing here is clinical guidance or a substitute for professional judgment.
Technical failure can become human consequence.
If you are introducing a capability into an environment where being wrong is measured in outcomes rather than error rates, that is the work we find most interesting.