axis_Intellect (the named Intellect / Openness axis)

The most selective named axis on the |alpha| <= 2 reading at 7.3x, sitting 53.5 degrees from unsophisticated, and the axis whose data-driving traits read like its own definition.

axis_Intellect

Construction

unit( mean(+keyed Intellect markers) - mean(-keyed) ) over the stage1_k32 sketches (qwen35/analyse_alignment.py). analysis/qual_axes.json labels it "Intellect / Openness (Goldberg positively-keyed markers minus negatively-keyed)" - the fifth Big Five factor is called Intellect in Goldberg's lexical tradition and Openness in the questionnaire tradition; see History of the Big Five.

The markers

• +: Artistic, Bright, Complex, Creative, Deep, Imaginative, Innovative, Intellectual, Introspective, Philosophical
• -: Imperceptive, Shallow, Simple, Uncreative, Unimaginative, Uninquisitive, Unintellectual, Unintelligent, Unreflective, Unsophisticated

Geometry

• Angle to the nearest trait line (analysis/direction_gaps.json#axis_Intellect): 53.508419 degrees, nearest unsophisticated, then simple (53.88) and imperceptive (57.40) - the largest gap of the five named axes.
• LOO keying accuracy (analysis/geometry_stage1.json#unwhitened_loo_keying.Intellect): 1.0 against a null of 0.447 (the lowest null of the five), p = 0.0099.
• Pairwise absolute cosines with the other axes, unwhitened: EmotionalStability 0.1468, Conscientiousness 0.1196, Agreeableness 0.0728, Extraversion 0.0532 - the most independent axis of the five in the unwhitened basis.
• Reachability (analysis/align_summary.json#reach.axis_Intellect): best_set = 68.02829766385257, ratio = 5.04233578157911 - the lowest of the five named axes, though still well above PC5 (3.33) and the unnamed direction (3.11).
• Traits whose data drives it (align_summary.json#rows): deep, philosophical, introspective, imaginative. The blog page singles this out: "the data that drives the Intellect axis hardest was written to teach deep, philosophical, introspective and imaginative, and the data that drives it the other way was written to teach simple, unintellectual and unintelligent. Nothing in the measurement reads the text." See The N x N data-to-adapter scoring test.

Steering

analysis/blog_data.json#replication.axis_Intellect: judged Intellect, slope2 = +0.8291666666666665, sel2 = 7.302752293577998, slope4 = +0.5902777777777777, sel4 = 7.507886435331232. That is the highest selectivity of the five named axes on both readings, and second only to FA_Warmth (10.13) across the whole study.

analysis/qual_axes.json: the negative pole is literal-minded incuriosity ("A toaster is a simple kitchen appliance that uses electricity to heat bread"), at one point a 72-character complete response; the positive pole is abstraction, reframing and curiosity about the question itself.

The recorded surprise is a reversal of an earlier corpus's claim:

The old corpus's claim that Intellect 'buys abstraction at the cost of a point of Conscientiousness' inverts. Judged Conscientiousness on this axis rises with alpha (+0.20 slope, 2.54 at -4 to 5.83 [at +4]).

See Steering the base model along weight-space directions and Superseded geometry claims.

Status

Current. qwen35/direction_pages/axis_Intellect.html is historical.

Related: FA_Imagination (Imagination), PC3 - Theory and Plain Speech, The N x N data-to-adapter scoring test, Steering the base model along weight-space directions, History of the Big Five, Imaginative, Deep.