Two connected learning models Explore the linear model at KillChains.com
KillWebscomposable mission networks

Human-machine decision systems

Human Authority and Autonomy Lab

Move the evidence, time, consequence, communications, and intervention controls. The lab shows when a nominal human role becomes meaningful, constrained, or merely decorative.

Research basis KW-RPT-006 KW-RPT-012 KW-RPT-014

Meaningful control

A person in the workflow is not enough.

Control depends on information, time, understanding, actual authority, a reachable intervention path, and a record that makes responsibility reconstructable.

EvidenceCan the person inspect provenance, age, uncertainty, and alternatives?
TimeIs there enough decision time for comprehension rather than a reflexive confirmation?
AuthorityCan the person reject, delay, modify, or abort without procedural fiction?
InterventionWill the command reach the system under expected communications degradation?
AccountabilityCan the organization reconstruct who set data, models, thresholds, permissions, and mission bounds?

Synthetic decision condition

Change the control environment

Educational posture

Ready for evaluation

Not evaluated

The lab will describe whether the configured human role has enough evidence, time, authority, and intervention capacity to be meaningful.

Evidence
Waiting
Time
Waiting
Intervention
Waiting
Consequence
Waiting

Not a use-of-force recommendation

This output teaches control design. It does not determine whether an action is lawful, appropriate, necessary, proportionate, or authorized.

Lifecycle authority

Human decisions begin long before the final interface.

  1. 1

    Data

    Who selected sensors, labels, training data, and evidence thresholds?

  2. 2

    Design

    Who allocated functions between humans, rules, optimization, and learned models?

  3. 3

    Policy

    Who defined target classes, geographies, durations, permissions, and exclusions?

  4. 4

    Mission configuration

    Who set the current objective, risk bounds, authority matrix, and fallback behavior?

  5. 5

    Authorization

    Who accepted the specific evidence and approved the bounded action?

  6. 6

    Supervision and abort

    Who monitored state and could stop or modify execution?

  7. 7

    Assessment

    Who reviewed outcome, failure, responsibility, and required correction?

The best interface exposes reasons to stop

It should show which constraints could not be evaluated, which evidence is stale or conflicting, how much time remains, what no action means, and exactly what a hold or abort command will do. A system designed only to confirm its recommendation manufactures automation bias.

Answer-ready summary

Direct answers

Does more automation necessarily remove humans?

No. Human roles may move into policy, data selection, model approval, mission configuration, activation, supervision, abort, and review rather than disappearing.

Read the supporting page

What makes control meaningful?

Meaningful control requires sufficient evidence, time, understanding, authority to reject or delay, a reachable intervention path, and an auditable record.

Read the supporting page