Karolinska eye-tracking SVM flags reading risk at 95.6% accuracy

AI system

System: Eye-tracking during reading with a support-vector-machine classifier (SVM-RFE, 48 features)
Model: Not assessed
Release: Not assessed
Application developer: Karolinska Institutet (Nilsson Benfatto, Öqvist Seimyr and colleagues); two authors hold equity in Optolexia
Role: Not assessed
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: Eye-tracking records of under one minute of reading. A labeled sample of 97 high-risk and 88 low-risk 9–10-year-olds. Feature selection (SVM-RFE) within repeated cross-validation.
Human oversight: Not assessed
Relation to AI: The classifier is the whole result. The authors describe it as a first step toward making eye tracking a viable screening method. No cited source shows it used in schools or affecting a child's support.
AI contribution: The classifier is the whole result. The authors describe it as a first step toward making eye tracking a viable screening method. No cited source shows it used in schools or affecting a child's support.

Outcome identifiers: Not assessed
Benefit: reached 95.6% (± 4.5%) accuracy with balanced sensitivity and specificity
Category: Improved Task Accuracy
Status: Not assessed
Beneficiary groups: People.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: reached 95.6% (± 4.5%) accuracy with balanced sensitivity and specificity
Comparison: chance-level classification on the same cross-validation (and random feature selection)
Results: Reference: Chance: 49.1% ± 13%. Result: 95.6% ± 4.5%.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Participants: Developer: Karolinska Institutet (Nilsson Benfatto, Öqvist Seimyr and colleagues); two authors hold equity in Optolexia. Operator: None in the sources (research study).
Jurisdictions: Developer: Sweden.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Conditions for replication: Eye-tracking hardware in the screening setting. Reading material and norms for the child's language. Validation on an independent sample, which the cited source does not report.
Replication attempts: Not assessed
Later findings: No later development was found; the Optolexia equity and VINNOVA funding are disclosed in the paper itself.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Assessed benefits: People.
Depth: Not assessed
Reasons: People: No individually attributable outcome: the source reports cross-validated classification accuracy (95.6%) on 185 children, not what followed for any child.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: The makers' claim

Kind: Not assessed
Related entries: Not assessed
Field: Education / learning-disability screening. Dyslexia risk screening of 9-10-year-olds.
Purpose: The method was developed to screen for dyslexia risk from eye movements. The classification accuracy is that purpose.
Dates: Date basis unassessed: 2016 (precision unassessed). Updated: 22 October 2026.
Stage: Not assessed
Version: SHA-256: 1a2a7defa892…

Peer-reviewed study Developer

Nilsson Benfatto M, Öqvist Seimyr G, Ygge J, Pansell T, Rydberg A, Jacobson C (2016). Screening for Dyslexia Using Eye Tracking during Reading. PLoS ONE 11(12): e0165508 (PMC5147795).
Peer-reviewed study Developer

Nilsson Benfatto M, Öqvist Seimyr G, Ygge J, Pansell T, Rydberg A, Jacobson C (2016). Screening for Dyslexia Using Eye Tracking during Reading. PLoS ONE 11(12): e0165508 (publisher page).