Mindspeller Neuroprofiling. How task performance, EEG, and association evidence support O*NET-based role guidance

1 Executive Summary

• Only task-and-EEG evidence creates occupational evidence; self-report and association evidence add context to the final report.The report assigns approximately 5% emphasis to self-report, 35% to association evidence, and 60% to task-and-EEG evidence; these are product settings, not reliability coefficients or acceptance rules.Source: §1 · p. 3
• Role confidence is capped at Moderate because pilot verification does not establish prediction of training success, job performance, retention, or hiring outcomes.The software checks show that the implemented process runs as intended, not that the profile predicts those outcomes.Source: §1 · p. 3

2.1 The Say–Do Discrepancy

• Differences between self-report and task behaviour are reported as alignment or tension, not automatically treated as error.They may reflect aspiration, context, or a genuine difference between self-perception and measured task behaviour.Source: §2.1 · p. 4

2.2 Remote Recruitment and Incomplete Evidence

• The Neuroprofile adds structured evidence to remote recruitment and can guide follow-up questions, but does not replace interviews, qualifications, work samples, or human judgement.Remote interviews and short assessments may provide only partial views of performance under multiple cognitive demands.Source: §1 · p. 3, §2.2 · p. 4

2.3 AI as a Pair Worker

• The profile does not measure AI productivity, but describes task patterns that may matter in AI-assisted work, including structured checking, anomaly detection, and switching.Other examples include comprehension and idea generation.Source: §2.3 · p. 4

2.4 Cognitive-Fit Matching Is Not Hiring Suitability

• Role matching compares measured cognitive abilities with occupational cognitive requirements; it is not a hiring decision or a measure of overall hiring suitability.A person may show cognitive fit yet lack job experience or qualifications, while an experienced candidate may still receive only Moderate confidence.Source: §2.4 · p. 4

2.5 Design Objective

• The design objective is to connect each occupational statement to supporting evidence while separating measured findings from context, caveats, and out-of-scope information.The final AI-generated text must remain within those limits.Source: §2.5 · p. 4

3.1 Evidence Streams and Processing Layers; 3.2 Current Measurement and Evidence Architecture; 3.4 Backend Integration Safeguards

• The task-and-EEG branch alone leads to O*NET abilities and roles; self-report and association branches add context before final synthesis.Most occupational decisions are rule-based, while AI components summarize or describe approved evidence and cannot create abilities or roles or override evidence checks.Source: §3.1 · p. 5, Figure 1 · p. 5, §3.3 · p. 6

4.1 Rational Self-Report; 4.2 Association-Based Semantic Positioning

• Self-report captures conscious goals and preferences but is affected by self-presentation and introspection limits, so it provides context rather than creating abilities or roles.Association responses position participants in a semantic network and may indicate motivational themes, but do not create O*NET abilities or roles.Source: §4.1 · p. 7, §4.2 · p. 7

4 Evidence Streams and Synthesis Policy; 4.3 Task-and-EEG Evidence

• Task-and-EEG evidence combines behavioural scoring of task performance with the neural state recorded during the cognitive demand.For some abilities, behavioural results carry the main evidential weight, with EEG providing supporting context rather than proving the ability alone.Source: §4.3 · p. 7

4.4 The 5/35/60 Synthesis Policy

• The report gives approximately 5% emphasis to self-report, 35% to association evidence, and 60% to task-and-EEG evidence; these weights shape narrative emphasis, not ability or role qualification.The values are not reliability estimates or probabilities that statements are correct and should be revisited after empirical validation.Source: §4.4 · pp. 7–8

5.1 Four-Electrode EEG Recording Configuration

• The four-electrode montage records Fp1 and Fp2 frontopolarly and O1 and O2 occipitally, but does not support precise source localization or comprehensive coverage of several brain regions.The sparse setup also cannot provide a full account of brain-network connectivity or direct motor-cortex measurement.Source: §5.1 · p. 8

5.2 Continuous-Block Acquisition

• Each task uses one uninterrupted scored EEG block to measure a sustained task state, not response-locked ERP components; incomplete or poor-quality blocks trigger a repeat.Speaking, typing, and button presses are delayed until after the scored period wherever the task allows.Source: §5.2 · p. 8

5.3 Sampling, Filtering, and Feature Families

• EEG is sampled at 512 Hz, filtered to approximately 1–45 Hz, and analysed in overlapping two-second windows; only selected feature families influence ability evidence.Gamma is withheld when the EMG check suggests facial-muscle contamination; peak descriptors are for technical reporting and are not mapped directly to abilities.Source: §5.3 · p. 8

5.4 Same-Eye-State Baselines and Task-Evoked Change

• Each task is compared with a baseline in the same eye condition, using the change from the participant’s matching baseline as the main feature.The within-person comparison reduces the influence of stable participant, device-gain, and electrode-contact differences.Source: §5.4 · pp. 8–9

5.5 Evidence Gating

• EEG features must pass signal-quality, change-direction, statistical-evidence, and practical-meaningfulness checks; EEG change alone cannot prove a stable trait, ability, brain source, or diagnosis.Multiple-comparison control and dependence-aware aggregation are used because many EEG features move together.Source: §5.5 · p. 9

6.1 How the Ability Set Was Chosen

• The review of all 52 O*NET abilities considered occupational discrimination, breadth of usefulness, measurability, risk of misleading labels, and information beyond abilities already selected.The instrument must be able to measure an ability without inviting a misleading claim.Source: §6.1 · p. 9

6.2 How Abilities Are Classified

• Abilities are classified as EEG/task-measurable, multimodal, caveat-only, or blocked; caveat-only abilities never improve a match, and blocked abilities cannot provide positive evidence.Source: §6.2 · pp. 9–10

6.3 Production Ability Space

• The 23-ability production set comprises 17 EEG/task-measurable and six multimodal abilities; Oral Expression and Speech Clarity are excluded because the battery and four-electrode recording cannot establish speech-production quality.Source: §6.3 · p. 10, Table 2 · p. 10

7.1 Design Constraints

• All twelve production tasks can contribute to role matching only when their behavioural and EEG checks pass.The tasks are designed for the four-electrode protocol, uninterrupted scored blocks, limited movement or responses during recording, and complementary evidence coverage.Source: §7.1 · p. 10

7.3 How a Task Earns an Ability: the Three-Field Contract

• Only behaviourally supported abilities with usable EEG that passes the evidence checks enter the ability pool for role matching.The three fields distinguish theoretical candidates, behaviourally validated candidates, and allowed candidates; design intent or an EEG change alone is insufficient.Source: §7.3 · p. 12

7.4 Graded Behavioural Evidence

• Speed and accuracy tasks use graded evidence, but pilot latency bands and cut-offs still require empirical calibration.Weaker grades can reduce evidence quality; strong grades do not add a bonus beyond a valid, unpenalized result.Source: §7.4 · p. 12

7.5 Rubric-Based Free-Text Scoring

• For free-text tasks 6 and 12, a language model scores fixed rubric dimensions, while deterministic rules—not the model—decide ability passes.Missing or malformed model output receives no score rather than a default pass.Source: §7.5 · p. 12

7.6 Session Progression and Parallel Forms

• Parallel task forms allow stability checks without reusing stimuli, while added tasks broaden ability coverage; more sessions alone do not raise confidence.Confidence rises only when repeated behaviour, EEG quality, cross-task support, and validation rules justify it.Source: §7.6 · p. 12

8.1 Why O*NET Is Used; 8.2 How Roles Enter the Bank

• O*NET release 30.3 supplies a reproducible occupational vocabulary; only requirements within the measurable 23-ability set are considered for role matching.Role curation weighs measurable coverage, representation, independent information, and the risk of overstating unmeasured requirements.Source: §8.1 · p. 13, §8.2 · p. 13

8.3 Role-Bank Composition

• The internal bank contains 376 occupations, and roles enter only when their measurable cognitive requirements meet the applicable bank rules; requirements outside the protocol remain caveats.Source: §8.3 · p. 13

8.4 Eligibility, Selection, and Role Confidence; 8.5 Portfolio Policy and Supplementary Pathways; 8.6 Role Matching Is Not Hiring Suitability

• Eligibility, portfolio selection, and role confidence are separate decisions; the report selects a small portfolio with an approximate 80/20 professional-or-knowledge-work to trades-or-field-operations target.A supplementary pathway illustrates a similar measured cognitive pattern, not job readiness; role matching does not decide hiring suitability.Source: §8.4 · pp. 13–14, §8.5 · p. 14

• The layer groups recurring abilities into work-style themes, including structured reasoning, monitoring, perceptual comparison, spatial transformation, comprehension, workflow switching, ideation, and quantitative handling.These themes summarize measured evidence rather than introducing new performance measures.Source: §9 · p. 14
• Workflow switching is represented by Time Sharing and Category Flexibility, while controlled ideation draws on Fluency of Ideas, Originality, and Category Flexibility.Source: §9 · pp. 14–15
• Quantitative handling draws on Mathematical Reasoning and Number Facility; a monitoring theme does not establish vigilance in a safety-critical workplace.Delivery-readiness and diligence-related proxies likewise do not measure delivery speed or establish work ethic.Source: §9 · p. 15

10.1 Four Different Confidence Concepts

• The framework distinguishes signal reliability, ability evidence quality, role confidence, and Neuroprofile Match Reliability; a large EEG change or coherent profile does not automatically make an ability or role high confidence.Source: §10.1 · p. 15

10.2 How Conservative Caps Work; 10.3 Where Caps Apply

• The system applies the lowest applicable confidence cap, so strong evidence in one part cannot cancel a serious limitation elsewhere.Caps can limit claims at multiple stages without altering raw feature statistics.Source: §10.2 · p. 15, §10.3 · p. 15

10.4 The Current Pilot Cap

• Pilot status caps role confidence at Moderate for every participant, even after a complete session with technically clean data.High confidence requires participant-sample calibration and a formal validation-status change.Source: §10.4 · pp. 15–16

10.5 What Is Already Implemented and Verified

• The software checks signal, behavioural, evidence, and role rules; 344 automated tests and one real Session 3 run verified that implemented components connected and behaved as specified.The run produced 22 of 23 abilities, scored 376 role candidates, created a compact portfolio, and respected the Moderate cap; this is functional verification, not evidence of score validity or work-outcome prediction.Source: §10.5 · p. 16

10.6 What Has Not Yet Been Established

• Repeat-session stability, agreement with established measures, cut-off appropriateness, rubric-rater agreement, device consistency, expert agreement, and incremental outcome information remain open empirical questions.Source: §10.6 · p. 16

10.6 What Has Not Yet Been Established; 10.7 How Confidence Can Increase

• Mindspeller does not claim validated prediction of job performance, validated hiring suitability, or superiority over established selection systems.A future outcome study may test whether task-and-EEG evidence adds information beyond self-report and association evidence; no such result is claimed here.Source: §10.6 · p. 16, §10.7 · p. 16

11.1 Where AI Is Used

• Ability and role qualification remain rule-based; AI assists with interpretation and report writing rather than creating or qualifying roles.The role matcher is deterministic, and the synthesis agent receives role results rather than creating them.Source: Table 6 · p. 17, §11.1 · p. 17

11.2 Compact Handoff

• The synthesis model receives a compact, structured handoff instead of raw EEG arrays or all 376 role records, making its output easier to audit.Reviewers can compare the handoff with the final text to check whether the model stayed within the supplied evidence.Source: §11.2 · p. 17

11.3 Final Neuroprofile Output

• The Neuroprofile presents multiple views of the same handoff, so repeated content across sections is not additional evidence.The report may include cognitive and motivational summaries, task-evidence themes, career domains, pathways, work environments, development suggestions, confidence statements, and chart data.Source: §11.3 · p. 17

12.1 The Battery as a Reference Protocol

• The twelve-task battery is a reference protocol: stimuli, difficulty, parallel forms, and task material can change, while new tasks require specified evidence rules.Changes must preserve the link between task, construct, behavioural result, EEG context, and any ability reported.Source: §12.1 · pp. 17–18

12.2 Compatibility with the Four-Electrode Recording Configuration

• New modules must suit Fp1/Fp2/O1/O2 recording; tasks dominated by speech, movement, fine motor execution, or uncovered neural systems need other sensors or should remain behavioural-only.Compatible candidates favor continuous low-movement states, standardized stimuli, matching eye-condition baselines, and delayed responses.Source: §12.2 · p. 18

12.3 Role-Bank Expansion Follows Ability Evidence

• Role-bank growth follows ability evidence: define and test the needed ability before adding a role, and expand the task battery when a central ability is unsupported.Once relevant abilities are supported, adding a role is mainly a bank calibration and coverage decision.Source: §12.3 · p. 18

12.4 Remaining Scope Gaps

• Speech-production and interpersonal constructs remain outside current scope without scored spoken production or a validated social-cognition task; physical, manual, body-control, and low-level sensory abilities are also excluded.Optional affective, body-awareness, and colour-preference modules may add context but cannot raise an occupational ability score without a validated evidence route.Source: §12.4 · p. 18

13 Future Work

• The next phase should prioritize calibration and validation over simply increasing the number of roles, and test the 5/35/60 narrative policy rather than treat it as an established optimum.The policy remains a product setting until tested.Source: §13 · pp. 18–19
• Pilot accuracy, speed, and rubric thresholds should be calibrated from participant distributions, while repeated sessions with parallel forms test behavioural and EEG-feature stability.The proposed repeated-session reliability test uses parallel forms.Source: §13 · p. 18
• Production tasks should be compared with established measures of mental rotation, perceptual speed, closure, sustained attention, divergent thinking, working memory, and speech-in-noise comprehension.Source: §13 · p. 19
• Independent human raters should be used to assess agreement with the free-text grader before rubric-dependent abilities move beyond exploratory reporting.Source: §13 · p. 19
• An independent rater should repeat the ability review, while redundancy analysis is refined using occupation-by-ability data.Source: §13 · p. 19
• Speech-in-noise delivery needs standardized acoustic level and timing across supported hardware, and remaining task-characteristic identifiers need alignment with the HR/work-style scorer.Source: §13 · p. 19
• External validity should be tested against appropriate training or work outcomes and for added information beyond existing methods; internal matches and O*NET ratings are not ground truth.The passage places this study after the measurement model is stable.Source: §13 · p. 19

14 Glossary

• The role matcher uses the session's allowed ability pool to score a curated 376-occupation bank and produce a short portfolio, with adjacent and supplementary pathways carrying narrower meanings.Supplementary pathways illustrate cognitive style rather than literal job recommendations.Source: §14 · pp. 19, 21, 22
• Evidence checks include behavioural, signal-quality, direction, statistical, and change-size criteria, alongside recording integrity, reliability, and task-to-baseline feature differences.Task convergence means support from more than one task or characteristic family.Source: §14 · pp. 19, 20, 22
• The glossary distinguishes functional verification from psychometric validation and defines the pilot profile, measurement scope, confidence cap, and narrative weights as boundaries on current claims.The 5/35/60 synthesis weights are product settings, not validation coefficients.Source: §14 · pp. 20–22
• Occupational evidence comes only from task-and-EEG evidence; cognitive fit means overlap with cognitive role requirements, not practical readiness or hiring suitability.A work-style proxy summarizes repeated ability evidence rather than directly measuring workplace performance.Source: §14 · pp. 20–22
• Ability terms separate a task's theoretical candidate from behavioural support, allowed evidence, and downstream qualification; a blocked inference applies only to its declaring task.Source: §14 · pp. 19, 21, 22

15 Conclusion

• Only scored tasks and EEG qualify O*NET abilities and occupational matches; self-report and association evidence explain conscious goals, preferences, and motivational themes.Source: §15 · p. 22
• Four electrodes, twelve tasks, 23 production abilities, and 376 occupations define the current method; roles appear only when task and EEG evidence meet relevant rules.Source: §15 · p. 22
• External reliability and validity studies remain necessary, so role confidence stays capped at Moderate while the system is calibrated and tested.Source: §15 · p. 22
• The output is intended to support clearer conversations about measured cognitive fit while leaving qualifications, practical skills, and hiring decisions to employers and candidates.Source: §15 · p. 22