Truace
Subscribe
The Index recomputed live from the record

What the evidence says.What the public feels.

The record holds 933 sourced gains and 770 sourced problems, averaging 68 and 66 on the index score. Readers have logged 22 public signals on the Pulse, which is kept apart and never counted as evidence.

933 gains
770 problems
Every sourced claim in the Index, one square each, shaded by the strength of its evidence. High Moderate Emerging

Ranks distinct AI gain and problem claims from the published record. Scores reward impact, independent source strength, scale, confidence, and recency.

1,703 results
Show filters and sorting

Download every matching row, not just this page:Export CSVExport JSON

AI gains · 933

149
HealthStableModerate evidence · 1 source

AI-driven diabetic retinopathy screening using ultra-widefield fundus images achieved a summary sensitivity of 85.0% and AUC of 0.870 in meta-analysis.

A systematic review and meta-analysis to February 9, 2025, evaluated artificial intelligence for diabetic retinopathy assessment using ultra-widefield color fundus images, which capture a larger retinal area without pupil dilation. Of 527 records, 17 studies were reviewed and four were meta-analyzed, all using Optos software, to estimate sensitivity and specificity for AI-driven screening.

Impact 30%
69
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
85

Updated Jul 25, 2026 · TRV-2026-0564

73Index score
150
HealthStableModerate evidence · 1 source

A fine-tuned MobileNetV2 model differentiated vitiligo from postinflammatory hypopigmentation with 94.88% accuracy and 0.9885 AUC while providing clinically meaningful ensemble explanations.

A diagnostic accuracy study published July 24, 2026 developed an interpretable deep learning framework to distinguish vitiligo from postinflammatory hypopigmentation, two conditions with similar depigmented lesions. Using 332 clinical images from King Abdullah University Hospital and public sources, a fine-tuned MobileNetV2 was evaluated with patient-wise 5-fold cross-validation.

Impact 30%
69
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
85

Updated Jul 25, 2026 · TRV-2026-0563

73Index score
151
ScienceStableModerate evidence · 1 source

DynStabNet E(3)-equivariant graph neural network predicts dynamical stability of semiconductor crystal candidates without explicit phonon calculations at inference, achieving 97% accuracy and cutting per-structure evaluation from hours to ~1 ms to accelerate large-scale screening.

Researchers developed DynStabNet, an E(3)-equivariant graph neural network that learns to predict whether crystal structures are dynamically stable from phonon-informed training data, avoiding explicit phonon calculations at inference. As of the July 2026 publication, the model was reported to reach 97% accuracy while reducing evaluation time per structure from several hours to about 1 ms.

Impact 30%
69
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
84

Updated Jul 22, 2026 · TRV-2026-0510

73Index score
152
SportsStableModerate evidence · 1 source

SVM-based model using 18 features across four dimensions predicted injury risk in 800 university football players with 95.6% accuracy and 99.2% ROC-AUC, with SHAP identifying stress, sleep and balance as top factors.

Researchers built an 18-feature model across basic information, training, fitness and lifestyle dimensions for 800 Chinese university football players from a Kaggle dataset, comparing 10 algorithms and finding SVM best at 95.6% accuracy with SHAP highlighting stress, sleep and balance as key predictors.

Impact 30%
69
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
84

Updated Jul 22, 2026 · TRV-2026-0476

73Index score

AI problems · 770

149
BusinessNewly addedModerate evidence · 1 source

Human evaluators discounted creativity when told a product was made by AI instead of a human, driven by the belief that generative AI exerts less effort.

Researchers examined how knowing a creator is AI versus human affects creativity judgments for organizational and commercial artifacts. Across four experiments with 2039 participants, they found people sometimes rated the same product as less creative when told it came from AI.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
100

Updated Oct 9, 2026 · TRV-2026-1317

69Index score
150
EducationNewly addedModerate evidence · 1 source

When used in passive AI-directed mode, ChatGPT left students neutral at the resolution stage of cognitive presence, creating a resolution gap in critical thinking development.

On September 11, 2025, a mixed-methods study reported on 40 students from a southwestern US university who used ChatGPT for learning tasks. Survey data showed mean improvements in triggering events, exploration, and integration, while resolution remained neutral, and script analysis identified passive AI-directed versus collaborative AI-supported interaction patterns.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
99

Updated Oct 8, 2026 · TRV-2026-1312

68Index score
151
HealthNewly addedModerate evidence · 1 source

The same models showed very low inter-model agreement and significant temporal instability, with Gemini performing worse in evening sessions, indicating they cannot replace clinician judgment for vertical root fracture decisions.

A peer-reviewed study in Odontology compared ChatGPT-3.5, ChatGPT-4o, and Google Gemini on 60 expert-validated true/false questions about longitudinal tooth fractures, querying each model three times daily for 10 days for 5,400 total responses against expert reference answers.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
99

Updated Oct 7, 2026 · TRV-2026-1305

68Index score
152
PolicyNewly addedModerate evidence · 1 source

LLMs can generate highly convincing misinformation that exploits audience biases, and exposure was found to reduce trust and influence decision-making.

A September 2025 scoping review in AI & SOCIETY synthesized 24 empirical studies on generative AI, particularly LLMs, examining how they generate, detect, mitigate, and impact misinformation.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
99

Updated Oct 6, 2026 · TRV-2026-1300

68Index score

Recomputed live from the record · Oct 11, 2026, 6:21 PM