What counts as an improvement, and for whom?
Are the criteria explicit enough to challenge? Whose outcomes are included, who bears the costs, and where do people disagree about what a better decision would achieve?
Research
Make the reasoning visible.
Then examine it.
Our research asks how decision engineering can help people examine objectives, evidence, constraints, alternatives, and tradeoffs. The starting question is simple: what counts as an improvement, and for whom?
Questions guiding the work
What should we
be able to show?
Can people inspect the reasons?
Can a person identify the evidence, assumptions, and constraints behind a finding, including where information is incomplete?
Can they compare alternatives?
Does the explanation connect each option to stated objectives and criteria, make tradeoffs visible, and show when more evidence might change the comparison?
Can they trace the effect of a change?
When an input or arrangement changes, can people identify which conclusions and follow-up work deserve another look?
What transfers between applications?
Which reasoning structures are reusable across domains, and which depend on the vocabulary, models, and workflows of a particular application?
Evaluation approach
Start with a concrete task.
Keep the claims testable.
These principles guide our planned evaluations: define a question, make a fair comparison, and document what the results can and cannot tell us.
01
Define improvement before evaluating it
Describe the decision, the people affected, and the criteria for evaluating outcomes. Record disagreements about those criteria and make clear whose perspective a measure represents.
Make the standard of comparison inspectable.
02
Make the comparison explicit
Use the same decision context to examine different ways of presenting the reasoning. Look for whether people can explain an alternative, challenge an assumption, or identify a tradeoff.
Document the evidence and limits of the comparison.
03
Look for failures as well as benefits
Examine where a model omits someone’s interests, an explanation conceals uncertainty, or a change fails to prompt review. Check whether an apparent benefit shifts effort or cost to someone else.
Treat limitations as part of the findings.
Publications coming soon. Future reports will include methods, findings, and limitations.
Participate
Bring a question.
Challenge an explanation.
We welcome conversations with researchers, practitioners, and people who review decisions in their own work. Useful starting points include a procedure change, a technical choice under uncertainty, or a disagreement about whose work a proposed improvement serves.
Tell us about the question you would like to examine and which application or decision context it concerns.
contact@hermodlabs.com