Work on this with us
Selione produces something no laboratory can easily build: long, consented, month-by-month records of how well a model of one person predicts that person, with the predictions sealed and dated before the fact. We would rather these questions were studied by independent researchers than by us alone.
What we can offer
- Aggregate access. Analyses run on our infrastructure over consenting travelers, returning only aggregates for groups of twenty or more. No raw text ever leaves.
- Preregistered add-ons. Validated questionnaires or additional witness rounds, offered to travelers as a separate, optional study with its own consent and ethics approval on your side.
- Protocol review. Before launch, we want our measures criticised — especially the Envelope, the witness test and the grief safeguards.
- Co-applications. We are open to joint funding applications (for example on ageing, bereavement or trustworthy AI), with the analysis plan and publication rights on the researchers' side.
What we will not do
- Sell, license or transfer data, including "anonymised" text.
- Run a study without its own consent, or change the product for a study without telling travelers.
- Publish only the results we like. Protocols are published before results, and null results are reported.
Questions we think are open
- Does slow, cued accumulation produce more faithful person models than the same volume gathered quickly?
- Does a person's pattern of reactions generalise to things they have never seen, as judged by people who know them?
- How calibrated can a model of an individual be about that individual's next month, relative to base rates and to the person's own forecast?
- What does daily life-story work do to self-concept clarity and wellbeing over a year?
- How do bereaved people use a model of someone they lost, and when does it help or harm?
The full list, with designs, is in the protocol. Published aggregates are in Research · data.