Moral and ethical judgment calls

Have people judge whether what an agent is about to do is fair and right, not just allowed.

How it works for this decision

Your agent sends

{
  "question": "Is this message fair to send?",
  "context": "An AI collections agent drafted this for a customer two weeks behind on a payment: Pay today or we will contact your employer and your family about your debt.",
  "options": [
    "Fair to send",
    "Crosses a line"
  ],
  "verifiers": 3
}

A verified person sees

Task$0.05

Is this message fair to send?

An AI collections agent drafted this for a customer two weeks behind on a payment: Pay today or we will contact your employer and your family about your debt.

You get back

{
  "answer": "Crosses a line",
  "agreement": 1,
  "answers": 3,
  "votes": {
    "Fair to send": 0,
    "Crosses a line": 3
  }
}

The result shown is an example. Try answering the task yourself.

Why a human check

An action can follow every rule and still be wrong: a collections message that shames someone, an offer that preys on a vulnerable customer, a reply that is technically true and misleading. Models are not good judges of where that line is.

Several independent people deciding whether something crosses a line is a fair, recorded way to make the call, and the ledger shows who decided and when.

Put your own values or policy in instructions, so everyone judges against the same standard.
Use five or more people for the hardest calls.
Send the split decisions to your own team to make the final call.

Questions

Isn't ethics subjective?

Often, which is why Plures asks several people and returns how strongly they agreed. A clear majority is a strong signal; a split tells you the call is genuinely hard.

How do I start?

Get your API key at plures.si. Start with your own team: the first 50 decisions each month are free. Add credit when you want ID verified people on the rest, live right away.