Active project

Predict
the Self

How much of a later self-authored self-description can be predicted from an earlier one?

01 / Question

Forecast expressed identity

The first study is Dr. Jason Jeffrey Jones’ Predict Future Selves challenge. It treats later personally expressed identity text as a prediction target and begins the larger inquiry: how predictable are human lives?

Status: 81 test predictions frozen; private score pending
Unit: Longitudinally paired self-descriptions
Current result: Complete 50-case public-development scorecard

02 / Framework

An ipseological task

A bio is personally expressed identity: language an individual publishes to describe themself. Its words, hashtags, abbreviations, and emoji may act as identity signifiers. The forecast concerns a later expression of identity, not a hidden or supposedly “true” self.

Read Ipseology

03 / First forecast

Stable signifiers, unstable text

A deterministic projection learned which words tended to recur across the 150 training pairs, retained high-persistence statements, and predicted response length. On 50 public development cases it modestly improved edit similarity, token Jaccard, and ROUGE-L over repeating 2024 verbatim, while reducing word-count error from 56.68 to 41.64 words.

The tradeoff matters: token-overlap and character n-gram F1 declined, line-count error increased, and predicted text remained much more similar to the earlier response than real follow-ups were. There is no composite score or overall winner. The 81-case test artifact passes the official validator but remains unscored because its answers are private.

Read the full report