How Eevo decides

A model trained to weigh what options lead to for the people they affect, one stakeholder at a time.

Eevo is a consequential decision model by Upright. It judges options by what they lead to: who is better or worse off, by how much, and how likely that is. It is a model trained for decisions, not a text generator, so it answers with a choice and a probability rather than prose.

One run per stakeholder

A decision usually makes some stakeholders better off and others worse off. Eevo runs once for each stakeholder you name, with the same context and options, and asks which option is best for that stakeholder. It never adds the stakeholders up into one verdict: how much each one counts is a value judgement, and you make it. What Eevo does show is where they agree and where they conflict.

Two ways to weigh

In Default mode, the one a request without mode runs in, you list the consequences, and Eevo sizes them against each other: it compares them in pairs for each stakeholder, solves one size per consequence, takes overlap between them into account, and adds them up into each option's expected value. It shows how it reached the answer, which is why it is the mode for supporting agentic and human judgement. In Direct mode Eevo works out the consequences itself, from the context, and answers with the best option and a probability for each, in one model call per stakeholder: faster, and suited to high-volume decisions and custom tasks.

Only from what it is given

Eevo knows nothing beyond the request. It does not look things up, and it cannot tell when a fact that would decide the case is missing. That is why its answers are information for a decision, not the decision itself, and why revisions exist: when a result shows something was left out, send the decision again with it.