Data and codebook

Two products, on different schedules and with different rules.

Aggregate release — weekly, open, no request needed

Distributions, means, standard deviations, internal consistency and the full correlation matrix, for every norm cell that survives suppression. Downloadable below. Everything on /explore is computed from exactly this file.

Item-level release — on request, consent-gated

One row per response: item id, item version, scored value, latency, revision count, presentation arm, ring order seed. Restricted to respondents who explicitly consented to the open research dataset, which is a separate checkbox from consenting to take part. Non-consenters are excluded from every release, permanently.

Age is stored in years and released in bands. IP is used at write time to derive a coarse region as a data-quality cross-check and then discarded, so there is no IP column to withhold. There is no name, no email in the main table, and no field that identifies a person.

These are applied before anything is written to a file, not at display time. A suppressed cell serialises as null; the client is never sent a value it is trusted not to show.

1. No cell below n = 25.
2. Complementary suppression, iterated to a fixed point. If exactly one child of a published parent were hidden, subtracting the published children from the parent would recover it exactly, so a second child is hidden too — and because that can expose a cell elsewhere, the process repeats until nothing changes.
3. At most 3 simultaneous filter dimensions, regardless of cell size. A large cell defined by many conjoined attributes is still a re-identification vector.
4. Weekly publication, counts rounded to the nearest 5, a cell refreshed only once it has gained 10 respondents, and no minimums, maximums or exact counts ever published. Without this, someone who knows roughly when a target responded can diff two releases and recover them exactly.

3.1 Session and demographics

3.2 Responses

3.3 Big Five scales

Thirty facets and five domains, from the public-domain IPIP-NEO-120. A facet is the sum of four items and ranges 4–20. A domain is the mean of its six facet sums, not the sum of its 24 items, and therefore also ranges 4–20 — summing the items gives a 24–120 range and silently invalidates every published comparison.

3.4 Closeness module

The novel part. 10 facets are re-administered at 4 ordered levels of closeness (Inner circle, Friends, Acquaintances, Strangers), with 30 unique statement families expanded across the rings. Each family appears exactly once per ring, which is the invariant that makes ring scores comparable at all.

Everything in this table is provisional. The closeness module is a research bet with no external norms, and if it fails its pre-registered falsification conditions we will say so on this site rather than quietly dropping it. See /why for those conditions and /methods for the scoring.

• aggregates.csv — one row per norm cell per scale: n, mean, SD, alpha, and the full count vector. Everything /explore draws is in here.
• correlations.csv — the 0-row facet correlation matrix, long format.
• Both files are rate limited to 30 requests an hour per caller and carry an ETag, so a conditional request costs nothing. They change at most weekly — please cache rather than poll. This database is shared with a production service whose stability comes first, which is also why these are served from a precomputed rollup and never touch it.
• item-level.csv.gz — by request. Email the address on /about with what you intend to do with it. We have never refused a request and do not plan to; the step exists so that there is a record of who holds a copy, which is what we promised respondents.

Why CC0 on the items rather than CC BY. A full form mixes about 120 public-domain IPIP items with about 30 of ours. Under CC BY, every adopter would have to keep track of which is which and carry a notice for one subset — and nobody would. We cannot relicense IPIP’s items, so CC0 on ours is the only route to an instrument with a single licence.

CC0 does not mean we expect to go uncited. Academic citation runs on norms and convenience, not on licence terms, which is why the citation block below is on this page rather than three clicks away.

The dataset

@misc{ticknor2026spcdata,
  author       = {Ticknor, Robert},
  title        = {Concentric Social Proximity Conditionality Dataset},
  year         = {2026},
  version      = {not yet published},
  howpublished = {\url{https://concentric.trovvy.com/data}},
  note         = {CC0 1.0}
}

The instrument

@misc{ticknor2026concentric,
  author       = {Ticknor, Robert},
  title        = {Concentric: A Closeness-Conditional Measure of {Big} {Five} Trait Expression},
  year         = {2026},
  howpublished = {\url{https://concentric.trovvy.com}},
  note         = {CC0 1.0 items and scoring keys}
}

The construct

@misc{ticknor2026spc,
  author       = {Ticknor, Robert},
  title        = {Social Proximity Conditionality: Measuring {Big} {Five} Trait Expression as a Function of Social Closeness},
  year         = {2026},
  howpublished = {\url{https://concentric.trovvy.com/why}}
}

Please also cite the source instrument, which is not ours:

• Goldberg, L. R. (1999). A broad-bandwidth, public domain, personality inventory measuring the lower-level facets of several five-factor models.
• Goldberg, L. R., et al. (2006). The International Personality Item Pool and the future of public-domain personality measures. Journal of Research in Personality, 40, 84–96.
• Johnson, J. A. (2014). Measuring thirty facets of the Five Factor Model with a 120-item public domain inventory. Journal of Research in Personality, 51, 78–89.

“NEO”, “NEO-PI-R” and “NEO PI-3” are trademarks of PAR, Inc. We cite IPIP-NEO-120 as a source instrument and use no PAR material.

The closeness-conditional construct was documented by Robert Ticknor in August 2026, and the repository history is the primary record. A timestamped preprint with a DOI will be deposited at launch — a DOI is what keeps a citation resolvable after a domain lapses, and it is the difference between a claim and a citable one.

The closest prior work is cited rather than avoided. /why names the paper most likely to falsify this and explains why we think the distinction holds.