Gen Fukuhara | Computational Psychiatry Researcher
NCNP Joint Research | Quantifying Interaction in Japanese Conversation | Developer of Turn Lens
One question: how can the interaction between people in ordinary conversation be measured, quantitatively and reproducibly?
Statement
I am Gen Fukuhara, a researcher in computational psychiatry (CPSY). I have worked on a single question: how can the interaction between people in ordinary conversation be measured, quantitatively and reproducibly?
Since November 2025 I have pursued this as a collaborative researcher at NCNP (National Center of Neurology and Psychiatry, Japan), serving as research lead on a joint study of quantitative indicators of interactional features in Japanese conversation. Working from the Corpus of Everyday Japanese Conversation (CEJC), I defined 19 interactional features across four categories, then examined their validity against Big Five personality scores produced by four LLM raters.
Method matters more to me than novelty. Ridge regression, permutation testing over 5,000 iterations, bootstrap variance analysis over 500 iterations, sensitivity analysis, confound control, and baseline conditions designed to falsify my own result. All five Big Five dimensions remain significant after Holm correction (r = 0.254–0.423), with conscientiousness highest (C, r = 0.423, corrected p = 0.0040). Extraversion sits just under the threshold (corrected p = 0.0490) and is treated as a borderline result. No single dimension is singled out as the headline finding; all five are treated on equal footing.
The measurement pipeline is now a tool called Turn Lens. Its development is the subject of an accepted poster at the 13th conference of the Japanese Society for Support of Adults with Developmental Disorders (Kochi, September 2026).
Research: Quantifying Interactional Features
Lead lea × NCNP joint research. Phenomena long described qualitatively in conversation analysis, reorganized into a feature set that can be measured quantitatively and reproducibly over a large corpus.
Timing (PG)
Speech ratio, pause length, response gaps, overlap rate, and variability of silence.
Fillers (FILL)
Filler occurrence rate and fillers per 100 characters.
Sequence organization (IX)
Repair initiation, repair after questions, yes/no responses, and lexical alignment.
Response type (RESP)
Response types following Japanese final particles, including aizuchi patterns.
(10 classical / 9 novel)
(conversation × speaker)
(66 conversations)
(IPIP-NEO-120)
Method
Every step below exists to doubt the result, not to strengthen its appearance.
- Data selection: 140 at-home dyadic conversations from CEJC were expanded into 378 conversation × speaker records; only the 120 records meeting all of ≥80 utterance pairs, ≥2,000 characters and ≥10 post-question pairs were retained (258 excluded).
- Leakage control: Ridge regression (α = 100) evaluated with 5-fold CV using a GroupKFold subject-wise split, guaranteeing that no speaker appears in both training and validation folds.
- Permutation testing: 5,000 iterations, applied to both the overall model correlation and each individual regression coefficient.
- Multiple comparisons: the five Big Five dimensions are treated as five independent tests, corrected with the Holm-Bonferroni method.
- Coefficient stability: 500 bootstrap resamples yield the SD and 95% CI of each coefficient. Only features whose CI excludes zero are treated as stable contributors.
- Confound control: Model A (19 interactional features) is compared against Model B (Model A plus gender and age, 21 variables). O, C, A and N retain significance under control, so the association does not reduce to demographic confounding.
- Sensitivity analysis: the regularization parameter was varied across α = 10 / 50 / 100 / 200 / 500 (C: r = 0.409–0.423). O, C, A and N stay significant across the whole range. Only E fails at α = 10 and 50, so its significance is explicitly flagged as depending on that choice.
- Cross-validation design comparison: subject-wise split (GroupKFold) is set against ordinary KFold. For O and A, KFold returns a higher r, consistent with optimistic bias from speaker leakage. Every reported figure uses the subject-wise split.
- Falsification baselines: condition 1 (with text) vs. condition 2 (summary only) vs. condition 3 (shuffled text, random seed 42). Under condition 3 all five dimensions lose significance, confirming it works as a negative control.
- Dose-response experiment: filler quantity in the conversation text was manipulated at ×0 / ×1 / ×3 to test whether the LLM's C score tracks it. No change was detected (ΔC = −0.012, 0.3% of the 0–4 scale), so filler measures are read as correlational associates rather than causal determinants of the scoring.
- Staged model comparison (exploratory): Stage 1 (demographics only) → Stage 2 (+ classical) → Stage 3 (+ novel). In this sample only N shows a significant increment (ΔR² = +0.069), so this analysis is not used to support the main claims.
Main Results
Permutation test results for predicting virtual Big Five scores from interactional features. All five dimensions survive Holm correction (r = 0.254–0.423), and the spread between dimensions is small.
| Big Five dimension | robs | p (corrected) | Sig. |
|---|---|---|---|
| C — Conscientiousness | 0.423 | 0.0040 | * |
| N — Neuroticism | 0.410 | 0.0072 | * |
| A — Agreeableness | 0.397 | 0.0072 | * |
| O — Openness | 0.337 | 0.0124 | * |
| E — Extraversion | 0.254 | 0.0490 | * borderline |
No headline dimension
All five dimensions are significant and r falls in a narrow band (0.254–0.423). Picking one of them as "the strongest association" would be a post-hoc choice, so all five are treated equally. For reference, inter-rater agreement is highest for C (r̄ = 0.699) and lowest for A (r̄ = 0.435); at the individual-rater level C, A, O and N are significant in 3 of 4 raters, E in only 1.
Contributing features differ by dimension
23 features were "concordant" — permutation p < 0.05 and a bootstrap 95% CI excluding zero. Counts range from 2 to 7 per dimension (A: 7, O: 6, C: 5, E: 3, N: 2) and the composition differs by dimension. Timing features contribute to all five; the newly defined sequence-organization and response-type features contribute in four of five. Nine of the 23 are novel features, and for N both concordant features are novel.
The negative control worked
When text–speaker pairings are shuffled (condition 3), all five dimensions lose significance (r = −0.179 to 0.201, p ≥ 0.587). Condition 1 leads by Δr = +0.195 to +0.589 across every dimension.
Features capture individual differences
13 features correlate significantly with age (overlap rate ρ = 0.500, speech ratio ρ = 0.447, filler occurrence ρ = 0.442). Speech ratio and pause length also relate to gender.
Turn Lens: A Conversation Measurement Tool
The measurement pipeline, packaged so that a person can run it on their own conversation. Scan, consent, record, visualize. Results are shown only on the user's own device.
For adults with developmental disorders, mismatches in conversational tempo, turn-taking and timing often contribute to interpersonal difficulty, yet what exactly is mismatched remains hard to see from the inside. Existing interaction measures presuppose expert analysis and are not designed for self-review.
Turn Lens addresses that gap. The aim is neither diagnosis nor advice, but to give a person and their conversation partner something concrete to look at together, without blame.
- Case
- Two conversations between the presenter and spouse (June 2 and July 20, 2026, with consent), recorded and analyzed automatically.
- Conv. A
- One speaker held 85% of the floor (85:15); response gaps ran 4.4–4.8× baseline (z = +7.45), prompting the realization of talking too much.
- Conv. B
- Filler rate stood out (z = +5.54); the radar profile changed clearly within the same pair. Overlap rate was 0.000 in both.
- New insight
- High filler use and complete absence of interruption were features the user had not anticipated.
- Scope
- Research use only, not a medical service. Short conversations inflate the numbers.
Gen Fukuhara1,* et al. (1Lead lea LLC / NCNP joint research)
The poster reflects the analysis as of August 2026. Its claims of "four of five dimensions significant" and "agreeableness highest at r = .449" were superseded by the manuscript revision of 7 September 2026 (table above): all five dimensions are significant, with conscientiousness highest at r = 0.423. See the table above for current values. Click to enlarge
Research Interests
Quantifying conversational interaction
Turn-taking, response latency, alignment. Treating qualitatively described phenomena as reproducible numbers.
Computational psychiatry and measurement for adult developmental support
What measurement can a person actually use in a support setting: indicators as material for reflection, not diagnosis.
Reproducible evaluation design
Leakage control, correction for multiple comparisons, sensitivity analysis. Designs that let a result be doubted.
LLMs as measurement instruments, and the validity limits of that use
What are LLMs actually reading? Establishing the usable range via inter-rater agreement, internal consistency, and negative controls.
Limitations
Separating what can be claimed from what cannot. The following limits apply as of now.
- Big Five scores are LLM estimates (virtual Big Five), not self-report measures. Convergent validity remains to be established.
- The analysis covers 120 records (66 conversations, 74 speakers) from at-home dyadic conversations in CEJC only. No claim of generalization across settings or relationships.
- 25 of 74 speakers appear in two or more conversations (59.2% of records), so within-speaker correlation may underestimate standard errors. Within-speaker stability (ICC) is future work.
- The novel features (IX, RESP) contribute at the coefficient level in four of five dimensions, but whether they significantly improve overall model performance could not be established in this sample (N = 120). The coefficient-level verdict is an exploratory screen.
- Extraversion is significant but sits just under the threshold (corrected p = 0.0490) and is therefore borderline. At α = 10 or 50 it fails to reach significance, so the result depends on the choice of regularization. Its concordant features are limited to three pause measures; indicators for volume, topic initiation and laughter are missing.
- In every dimension the predictions are compressed: the fitted slope is shallower than y = x, so the predicted range is narrower than the observed range.
- In the baseline check, condition 2 (summary only) was unexpectedly significant across all five dimensions. Cronbach's α dropped below 0.3 there, which we read as the LLM leaning heavily on surface quantities when the input is impoverished.
- The Turn Lens case study covers two conversations from a single pair (N = 1 pair) and does not generalize. Research use only, not a medical service.
- This work is positioned as a proposal of quantitative indicators, not as building a personality-prediction model. The Big Five analysis is a means of validating construct validity.
Papers & Presentations
Development and case report of Turn Lens, a conversation visualization tool based on quantitative interaction indicators
13th Conference of the Japanese Society for Support of Adults with Developmental Disorders, Kochi (accepted July 28, 2026)
Conference detailsQuantitative indicators of interactional features in Japanese everyday conversation: proposal and validity assessment
A systematic proposal and validation of interactional features for a Japanese conversation corpus. 47 pages, 8 figures and 14 tables, with reproducibility scripts (revision of 7 September 2026).
Lead lea × NCNP joint research: quantitative indicators of interactional features in Japanese conversation
Approved October 29 and formally contracted November 4, 2025. Research lead: Gen Fukuhara (Lead lea LLC), with collaborators at the NCNP National Institute of Neuroscience.
Background
The research began outside my day job. Years of building and operating systems inform how I design measurable indicators.
2026.09
Revised the manuscript (47 pages), adding confound control, a cross-validation design comparison and a dose-response experiment; all five Big Five dimensions now show significant associations. Poster presentation on Turn Lens at the 13th JSSADD conference, Kochi.
2026.07
Completed the first manuscript draft and reproducibility scripts; poster accepted July 28.
2026.02
Appointed Executive Officer and CAIO, Prime AI (Vision Consulting).
2025.11
Became a collaborative researcher at NCNP, leading the joint study on interactional features.
2025.06
Selected for the AWS Japan generative AI implementation program.
2025.03
Launched Kankyo AI®, an emotion-analysis XAI service; core technology filed for patent.
2024.06
Spoke at AWS Summit Japan 2024 on automating surf-report generation with Amazon Bedrock (80% cost reduction).
2023.03–2025.03
CTO at First Trade Inc., leading healthcare SaaS and generative AI adoption.
Get in Touch
I welcome collaboration, data sharing and replication.
Questions about feature definitions or reproducibility scripts are equally welcome.
Gen Fukuhara
CTO, Lead lea LLC / Collaborative Researcher, NCNP / Executive Officer & CAIO, Prime AI
461-5 Kokufu-Hongo, Oiso, Naka-gun, Kanagawa 259-0111, Japan