Michael Mullarkey

Why they matter

Tracked through public AI activity and peer connections inside the directory.

AI signals
12
past 30d
Sources
1
distinct domains
Discussões
1
past 30d
Latest signal
9d ago
View every signal from Michael Mullarkey →
Clinical psych PhD | ML Engineer in sports analytics | he/him My newsletter on building cool data stuff: https://buttondown.com/datadash Personal website: mcmullarkey.github.io Opinions are my own and don't reflect the opinions of my employer

Articles & links

↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
  • A fine-tuned language model scanned 4,452 items across 459 scales in the International Personality Item Pool and flagged 504 jingle-jangle fallacy candidates.
  • Clustering compressed 277 construct labels into 68, cutting fallacies by 58% while requiring 75% fewer unique names.
  • The embeddings predicted scale correlations within mean absolute errors of 0.037 to 0.047 and fully recovered 81% of the original scales.
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
  • A fine-tuned MPNet model flagged 504 fallacy candidates across 4,452 psychometric items: 340 jingle (same label, different content) and 164 jangle (different labels, same content).
  • The authors propose a replacement taxonomy of 68 clusters that reduces fallacy candidates by 58% while using 75% fewer construct labels.
  • Embeddings predicted Cronbach's α with mean absolute error of 0.026–0.029 in-sample and 0.065–0.070 out-of-sample across 459 scales covering 277 constructs.
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
  • Dirk Wulff and Rui Mata fine-tuned an MPNet embedding model on 200,000 personality item pairs and vectorized 4,452 items across 459 IPIP scales.
  • The pipeline flagged 504 'jingle-jangle' fallacies: 340 scales sharing a label but measuring different things, 164 measuring similar things under different labels.
  • A reclustering pass cut fallacies by 58% while shrinking unique construct labels by 75%, and reproduced 81% of original scale structures.
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
  • A fine-tuned MPNet embedding model flagged 504 fallacy candidates across 459 scales in the International Personality Item Pool, split into 340 jingle and 164 jangle.
  • The embeddings predicted internal consistency with Pearson r = 0.75 in-sample and r = 0.61 out-of-sample across 4,452 IPIP items.
  • Letting the embeddings regroup constructs cut jingle-jangle fallacies by 58% and compressed 254 unique labels down to 68.
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
  • Wulff and Mata ran 4,452 items from the International Personality Item Pool through a fine-tuned language model and flagged 504 taxonomic fallacies: 340 jingle, 164 jangle.
  • The method's reorganized mapping produced 58% fewer jingle–jangle fallacies while requiring 75% fewer unique construct labels than the source taxonomy.
  • Embeddings predicted empirical internal consistency at r = 0.75 in sample and r = 0.61 out of sample, with 81% of 443 scales recovered in sample.
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
  • A Nature Human Behaviour paper by Wulff and Mata uses language-model embeddings to detect 504 candidate fallacies across 277 constructs in the International Personality Item Pool.
  • Embedding similarity predicted internal consistency at r = 0.75 in-sample and r = 0.61 out-of-sample across 449 scales, with 81% showing full structural recovery.
  • Clustering collapses the IPIP's 277 construct labels into 68 and yields a 58% reduction in jingle-jangle fallacies.
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
  • The method flags 504 candidate fallacies in the IPIP: 340 jingle and 164 jangle, across 4,452 items and 459 scales.
  • Predicted internal consistency tracks observed consistency at r = 0.75 in-sample and r = 0.61 out-of-sample across hundreds of psychometric scales.
  • A 68-cluster taxonomy cuts jingle-jangle fallacies by 58% while using 75% fewer construct labels than the IPIP's 277.
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
  • Wulff and Mata ran 4,452 items across 459 scales in the International Personality Item Pool through a fine-tuned language model to detect taxonomic fallacies.
  • The embeddings flagged 504 jingle-jangle fallacy candidates: 340 jingle and 164 jangle; a 68-cluster relabeling cut fallacies by 58% using 75% fewer labels.
  • Out-of-sample, predicted internal consistencies correlated with observed at r = 0.61, and 79% of 443 scales were fully recovered.
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
Read full analysis →
View on Bluesky →
↻ Michael Mullarkey reposted
Eiko Fried @eikofried.bsky.social

2025 paper on automated detection of jingle-jangle based on LLM embeddings. Here a proof of principle where they highlight personality scales high in label similarity (similar scale name) but low in item similarity [blue, jingle] and vice versa [beige, jangle]. www.nature.com/…

Semantic embeddings reveal and address taxonomic incommensurability in psychological measurement - Nature Human Behaviour nature.com
AI Weekly's analysis →
  • Wulff and Mata's embedding pipeline flagged 504 taxonomic fallacies in psychometric scales: 340 jingle, 164 jangle.
  • Applied to the International Personality Item Pool, embeddings collapsed 277 construct labels to 68 and cut fallacies by 58%.
  • The embeddings predicted internal consistency at r=0.75 in-sample and r=0.61 out-of-sample across hundreds of scales.
Read full analysis →
View on Bluesky →

Recent commentary

Was great to meet a bunch of folks at Posit conf! Here's me pictured with all the new lightbulb ideas I'm having from people's talks + convos (Not AI, it's a super cool exhibit at Houston's Museum of Fine Art called Aftermath of Obliteration of Eternity)

View on Bluesky · ♥ 13 ↻ 0 ↩ 1 · 64d ago

"If you don't talk to people, AI is a great way to automate misunderstanding" - @tcash21.bsky.social A theme across talks: building natural language tools exposes ambiguities that can otherwise hide Aligning on definitions + standardizing processes matters more now than ever #positconf2026

View on Bluesky · ♥ 7 ↻ 3 ↩ 1 · 66d ago

In Michael Mullarkey's orbit

Center = Michael Mullarkey. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.

Are you Michael Mullarkey? Show it.

Add the Who’s Who of AI badge to your site or bio. It links back to this profile.

Listed in AI Weekly's Who's Who of AI

Markdown: [![Listed in AI Weekly's Who's Who of AI](https://aiweekly.co/modules/custom/aiweekly_whoswho/images/whoswho-badge.svg)](https://aiweekly.co/whos-who/person/mcmullarkey-bsky-social)