IA on the job: should you ask for a table or a chart?
An assistant can turn production readings into a table or a chart, but these formats do not make the same operations equally easy. The management question is specific: which format should an AI be asked to produce so that a team can answer a production question correctly?
What the literature establishes
Iris Vessey (1991, Information Systems Research) proposes a theory of cognitive fit: the effectiveness of a representation depends on how well it matches the task. Tables are particularly suited to looking up precise values, while charts support certain spatial comparisons and the perception of relationships. The issue is therefore not which format is generally superior, but which supports the required operation.
What a representation makes directly visible
Jill H. Larkin and Herbert A. Simon (1987, Cognitive Science) show, through an analysis of problem-solving processes, why a diagram can facilitate certain inferences. Grouping related information spatially can reduce search and make relationships directly perceptible. Applied to AI, this mechanism suggests requesting a representation that makes the relevant relationship visible, rather than simply asking for more attractive formatting.
The premature conclusion
One might conclude that monitoring tables should be replaced with automatically generated charts. That would overlook the fact that the same set of measurements can be used to detect drift or to check an exact value for a batch: these tasks do not necessarily call for the same format. A presentation useful for examining a trend is not always sufficient for determining conformity. (our executive and employee training programmes)
What these studies do not establish
These articles address information representations and problem solving, not today's generative assistants. They establish neither that an AI-produced chart is accurate nor that it improves an industrial decision. The integrity of values, units, scales and tolerances must therefore be checked before assessing the usefulness of the format.
A practical check in Neuchâtel
As part of IA on the job, a Neuchâtel team working in microtechnology, watchmaking or precision microelectronics could compare a table and a chart produced from the same non-sensitive dimensional measurements, after checking that both accurately reproduce the data. Using comparable datasets and alternating formats between participants, ask them to find an exact value and then identify a drift defined in advance. For each task and format, measure the proportion of correct answers and response time separately, using an answer key prepared by the quality manager. The choice of format would then rest on observable performance, without changing batch-release procedures. To go further: explore the AI on the Job training in Neuchâtel in the canton of Neuchâtel, or browse our executive and employee training programmes in Switzerland.
In pictures: AI on the Job in Neuchâtel
- ai on the job
- Neuchâtel
- research
- training neuchâtel