Isaac Fradkin
@fradkinisaac.bsky.social
Assistant Professor at HUJI. Studying computational psychopathology, thought dynamics, disorders of thought and communication, semantic cognition and alignment, free association
https://sites.google.com/view/thought-dynamics-lab/
https://sites.google.com/view/thought-dynamics-lab/
(7/7) We are incredibly proud of this paper—the first in our lab's new journey into semantic alignment! Massive kudos to Siham Abu-Ali and @tanya-philippova.bsky.social for their fantastic work. Check it out here: osf.io/preprints/ps...
OSF
osf.io
June 15, 2026 at 9:50 AM
(7/7) We are incredibly proud of this paper—the first in our lab's new journey into semantic alignment! Massive kudos to Siham Abu-Ali and @tanya-philippova.bsky.social for their fantastic work. Check it out here: osf.io/preprints/ps...
(6/7) Thus, semantic misalignment may reflect a socially motivated modulation of identity rather than a mere failure of coordination. This may shed light on mechanisms of interpersonal distancing, communication breakdowns in neurodivergence, and broader patterns of social polarization
June 15, 2026 at 9:50 AM
(6/7) Thus, semantic misalignment may reflect a socially motivated modulation of identity rather than a mere failure of coordination. This may shed light on mechanisms of interpersonal distancing, communication breakdowns in neurodivergence, and broader patterns of social polarization
(5/7) In another experiment, we made alignment even simpler: we repeated cues, and showed participants the associations their partner (who tended to repeat their own associations) previously reported for a given cue. Again, we found less motivation to align with atypical partners.
June 15, 2026 at 9:50 AM
(5/7) In another experiment, we made alignment even simpler: we repeated cues, and showed participants the associations their partner (who tended to repeat their own associations) previously reported for a given cue. Again, we found less motivation to align with atypical partners.
(4/7) Our results show a fascinating dynamic! One possible strategy to maximize alignment is to produce more typical associations. Whereas people produce more typical associations in interaction, *they do so less when interacting with an atypical or difficult-to-align-with partner*.
June 15, 2026 at 9:50 AM
(4/7) Our results show a fascinating dynamic! One possible strategy to maximize alignment is to produce more typical associations. Whereas people produce more typical associations in interaction, *they do so less when interacting with an atypical or difficult-to-align-with partner*.
(3/7). But what happens when your conversation partner produces highly incoherent responses? In a new preprint, led by the brilliant Siham Abu-Ali, we tested this across four experiments using a novel word association task, where dyads generate associations to shared cues.
June 15, 2026 at 9:50 AM
(3/7). But what happens when your conversation partner produces highly incoherent responses? In a new preprint, led by the brilliant Siham Abu-Ali, we tested this across four experiments using a novel word association task, where dyads generate associations to shared cues.
(2/7). Conversations are usually about building common ground. We typically consider shared meaning to be the normative trajectory of linguistic interaction. More specifically, social interactions usually motivate us to produce conventional associations to align with others.
June 15, 2026 at 9:50 AM
(2/7). Conversations are usually about building common ground. We typically consider shared meaning to be the normative trajectory of linguistic interaction. More specifically, social interactions usually motivate us to produce conventional associations to align with others.
(8/8) We hope this work helps bridge NLP in psychiatry and theory-driven computational modeling by highlighting the potential use of simulations to enhance understanding.
A huge thanks to @schizbulletin.bsky.social, the editor, and the insightful reviewers for their pivotal input.
A huge thanks to @schizbulletin.bsky.social, the editor, and the insightful reviewers for their pivotal input.
November 30, 2025 at 12:28 PM
(8/8) We hope this work helps bridge NLP in psychiatry and theory-driven computational modeling by highlighting the potential use of simulations to enhance understanding.
A huge thanks to @schizbulletin.bsky.social, the editor, and the insightful reviewers for their pivotal input.
A huge thanks to @schizbulletin.bsky.social, the editor, and the insightful reviewers for their pivotal input.
(7/8) These new 'semantic density metrics' were also better at detecting repetitiveness in a reanalysis of a previous dataset, particularly for psychopathological dimensions corresponding with depressive and internalizing symptoms more generally. An interesting finding for future work!
November 30, 2025 at 12:28 PM
(7/8) These new 'semantic density metrics' were also better at detecting repetitiveness in a reanalysis of a previous dataset, particularly for psychopathological dimensions corresponding with depressive and internalizing symptoms more generally. An interesting finding for future work!
(6/8)
Findings: Cosine distance metrics were suboptimal in detecting simulated perseveration. They were outperformed by new semantic density metrics using dimensionality reduction to examine (roughly speaking) if a smaller number of sentences could have conveyed the same message.
Findings: Cosine distance metrics were suboptimal in detecting simulated perseveration. They were outperformed by new semantic density metrics using dimensionality reduction to examine (roughly speaking) if a smaller number of sentences could have conveyed the same message.
November 30, 2025 at 12:28 PM
(6/8)
Findings: Cosine distance metrics were suboptimal in detecting simulated perseveration. They were outperformed by new semantic density metrics using dimensionality reduction to examine (roughly speaking) if a smaller number of sentences could have conveyed the same message.
Findings: Cosine distance metrics were suboptimal in detecting simulated perseveration. They were outperformed by new semantic density metrics using dimensionality reduction to examine (roughly speaking) if a smaller number of sentences could have conveyed the same message.
(5/8)
To tackle this, we used generative language modeling to simulate texts characterized by derailment, repetitiveness, or both, and tested whether different NLP metrics could accurately capture these manipulations.
To tackle this, we used generative language modeling to simulate texts characterized by derailment, repetitiveness, or both, and tested whether different NLP metrics could accurately capture these manipulations.
November 30, 2025 at 12:28 PM
(5/8)
To tackle this, we used generative language modeling to simulate texts characterized by derailment, repetitiveness, or both, and tested whether different NLP metrics could accurately capture these manipulations.
To tackle this, we used generative language modeling to simulate texts characterized by derailment, repetitiveness, or both, and tested whether different NLP metrics could accurately capture these manipulations.
(4/8)
This leads to the key questions: Should we predict larger or smaller cosine distances in psychosis? What happens when derailment and repetitiveness in language co-occur?
This leads to the key questions: Should we predict larger or smaller cosine distances in psychosis? What happens when derailment and repetitiveness in language co-occur?
November 30, 2025 at 12:28 PM
(4/8)
This leads to the key questions: Should we predict larger or smaller cosine distances in psychosis? What happens when derailment and repetitiveness in language co-occur?
This leads to the key questions: Should we predict larger or smaller cosine distances in psychosis? What happens when derailment and repetitiveness in language co-occur?
(3/8)
But here’s a conundrum: Early findings correlated psychosis with greater cosine distances between utterances, yet recent studies found the opposite. Reduced distances may reflect repetitive language, common in psychosis, but also in other conditions, like depression🤷♂️
But here’s a conundrum: Early findings correlated psychosis with greater cosine distances between utterances, yet recent studies found the opposite. Reduced distances may reflect repetitive language, common in psychosis, but also in other conditions, like depression🤷♂️
November 30, 2025 at 12:28 PM
(3/8)
But here’s a conundrum: Early findings correlated psychosis with greater cosine distances between utterances, yet recent studies found the opposite. Reduced distances may reflect repetitive language, common in psychosis, but also in other conditions, like depression🤷♂️
But here’s a conundrum: Early findings correlated psychosis with greater cosine distances between utterances, yet recent studies found the opposite. Reduced distances may reflect repetitive language, common in psychosis, but also in other conditions, like depression🤷♂️
(2/8)
With the rise of Language Models, NLP in psychiatry is rapidly growing. Cosine distance metrics, for instance, are widely used to capture derailment and other forms of incoherence in conditions like psychosis.
With the rise of Language Models, NLP in psychiatry is rapidly growing. Cosine distance metrics, for instance, are widely used to capture derailment and other forms of incoherence in conditions like psychosis.
November 30, 2025 at 12:28 PM
(2/8)
With the rise of Language Models, NLP in psychiatry is rapidly growing. Cosine distance metrics, for instance, are widely used to capture derailment and other forms of incoherence in conditions like psychosis.
With the rise of Language Models, NLP in psychiatry is rapidly growing. Cosine distance metrics, for instance, are widely used to capture derailment and other forms of incoherence in conditions like psychosis.