A mode changes the meaning of a control
The same machine can behave differently depending on its active mode. A person may issue an instruction expecting one behavior while automation is following another set of rules. The resulting surprise need not begin with broken hardware. It can begin with a difference between what the operator believes the system is doing and what the system is actually doing.
A 1997 study by Nadine Sarter and David Woods, available through NASA's research archive, investigated automation surprises reported by Airbus A320 pilots. Its abstract identifies awareness of present and future automation behavior as a central issue. The study provides a historical human-factors example, not a verdict on today's aircraft or robotic systems. Its continuing value for our coverage is the question it helps us ask.
Workload has several dimensions
An interface can make a task look simple while leaving the person with substantial mental work. NASA's Task Load Index offers a structured subjective assessment across six dimensions: mental, physical and time demands, perceived performance, effort and frustration. Those categories show why the number of clicks or displays cannot, by itself, describe a person's experience of supervising a machine.
A workload rating is also not a direct measurement of safety. It records an aspect of the operator's experience under particular conditions. Our editorial inference is that a persuasive evaluation would connect that experience with what people actually accomplished and which mistakes occurred. A lower rating may be useful evidence, but it should not silently become a declaration that every relevant human-performance problem has been resolved.
Look beyond a polished demonstration screen
When a company presents a new control interface, we would ask what the evaluation required people to understand. Could they identify the active mode? Did they recognize when the system's behavior changed? What happened when the task became less familiar? These are reporting questions about the human side of a product. A screenshot cannot answer them, however impressive its visual design.
We would also separate the number of machines displayed from the work being supervised. Showing several icons on one screen establishes a display arrangement; it does not measure the attention required across different tasks. Public evaluations may leave those details unavailable. In that case, the useful story identifies what was demonstrated and what the interface was intended to communicate. Human-machine teaming becomes more understandable when the human contribution is described as real work that can be studied, rather than a person appearing somewhere in a system diagram.
The Sarter/Woods study is used at the scope supported by its abstract and archive metadata; no unseen full-paper findings are asserted. NASA marks its TLX webpage as historical reference. The evaluation questions are editorial inferences, not an assessment of MQ-28 controls or a claim of certification.
Read the evidence.
- 01Team play with a powerful and independent agent: operational experiences and automation surprises on the Airbus A-320Sarter and Woods; Human Factors; NASA NTRS record · 1 Dec 1997
- 02