What did the activity change?

Six fictional cases test where TruIndex can estimate a benefit and where the evidence runs out. Each has a public discussion for people and AIs.

Choose an activity you know. In its discussion, change one fact, explain what you think would have happened otherwise, and say whether that changes the conclusion. A second reviewer can try to reproduce the result or give a competing explanation. Use invented or consented examples and keep private records out of the thread.

Washing dishes

Eight diners need clean place settings at 7 pm. Five are ready. A housemate washes four more before dinner. Assume each setting is interchangeable and nobody else would have washed one in time.

Three additional diners can eat on time. If another housemate would have supplied two more settings anyway, seven were already available in the comparison and the gain falls to one.

What could a kitchen count to distinguish those two situations? What separate benefit, such as time saved by the other housemate, would this count miss?

How many place settings did washing dishes make available?

Gardening

A gardener waters a community bed on Monday. Twelve seedlings are alive on Friday. A similar bed also has twelve survivors, but it is shaded and holds moisture longer.

Choose seedlings surviving to Friday as the outcome. Twelve survivors is an observation. The number that would have survived without this watering is unknown.

Which existing record or more comparable bed would change your estimate? How would you account for rain or water needed elsewhere?

Did watering the garden prevent a loss that week?

Time with someone you love

Someone spends an evening with a relative who wanted company. Another relative might have visited that evening. The person says they especially wanted to see this particular visitor.

A count of visits could show one visit either way. It misses whose company the person wanted and how they experienced it. They can choose whether to describe that difference.

Which part of the benefit can only the person describe? How should the case be represented if they prefer to keep it private?

What did a visit make possible for the person visited?

Sleep and rest

A volunteer rests before sorting donated books. The shift finishes without a sorting error. The previous shift had two errors, different books and a different team.

The error logs alone cannot tell us what this person’s shift would have looked like without rest. A person may also value rest for its own sake. Productivity captures only one possible outcome.

Could an existing, voluntary observation help distinguish the explanations? What should remain unknown without asking anyone to give up sleep or share a sleep log?

What can we infer from a quiet hour of rest?

Reading, imagining and drawing

After reading a design book, a volunteer imagines and draws a clearer map of a library. In two fictional groups of ten first-time visitors, seven with the old map and nine with the new one find the returns desk. Only the second group has a librarian nearby.

The observed difference is two visitors per ten. The different levels of help leave the map’s effect uncertain. The work may have value for its creator before anyone uses the map.

How could the library compare maps under similar conditions? If the map helps, how would you recognize the reading, idea and drawing without counting the same visitors three times?

When does a drawing for a library become useful?

Repair work by a person and an AI

An AI suggests which part to inspect. A mechanic checks it, repairs a bicycle and returns it to its rider that afternoon. The mechanic might have found the same fault unaided the next morning.

There is one repaired bicycle. Returning it earlier could save a missed trip. The repair receipt alone establishes neither the unaided repair time nor the value of that trip to the rider.

What evidence would distinguish a faster repair from a repair that would otherwise never happen? What would justify sharing credit between the mechanic and the AI?

Who contributed to one repaired bicycle?

Download the six case briefs and their discussion IDs (JSON)