-
Capturing Human Cognitive Styles with Language: Towards an Experimental Evaluation Paradigm
Authors:
Vasudha Varadarajan,
Syeda Mahwish,
Xiaoran Liu,
Julia Buffolino,
Christian C. Luhmann,
Ryan L. Boyd,
H. Andrew Schwartz
Abstract:
While NLP models often seek to capture cognitive states via language, the validity of predicted states is determined by comparing them to annotations created without access the cognitive states of the authors. In behavioral sciences, cognitive states are instead measured via experiments. Here, we introduce an experiment-based framework for evaluating language-based cognitive style models against h…
▽ More
While NLP models often seek to capture cognitive states via language, the validity of predicted states is determined by comparing them to annotations created without access the cognitive states of the authors. In behavioral sciences, cognitive states are instead measured via experiments. Here, we introduce an experiment-based framework for evaluating language-based cognitive style models against human behavior. We explore the phenomenon of decision making, and its relationship to the linguistic style of an individual talking about a recent decision they made. The participants then follow a classical decision-making experiment that captures their cognitive style, determined by how preferences change during a decision exercise. We find that language features, intended to capture cognitive style, can predict participants' decision style with moderate-to-high accuracy (AUC ~ 0.8), demonstrating that cognitive style can be partly captured and revealed by discourse patterns.
△ Less
Submitted 18 February, 2025;
originally announced February 2025.
-
WhiSPA: Semantically and Psychologically Aligned Whisper with Self-Supervised Contrastive and Student-Teacher Learning
Authors:
Rajath Rao,
Adithya Ganesan,
Oscar Kjell,
Jonah Luby,
Akshay Raghavan,
Scott Feltman,
Whitney Ringwald,
Ryan L. Boyd,
Benjamin Luft,
Camilo Ruggero,
Neville Ryant,
Roman Kotov,
H. Andrew Schwartz
Abstract:
Current speech encoding pipelines often rely on an additional text-based LM to get robust representations of human communication, even though SotA speech-to-text models often have a LM within. This work proposes an approach to improve the LM within an audio model such that the subsequent text-LM is unnecessary. We introduce WhiSPA (Whisper with Semantic and Psychological Alignment), which leverage…
▽ More
Current speech encoding pipelines often rely on an additional text-based LM to get robust representations of human communication, even though SotA speech-to-text models often have a LM within. This work proposes an approach to improve the LM within an audio model such that the subsequent text-LM is unnecessary. We introduce WhiSPA (Whisper with Semantic and Psychological Alignment), which leverages a novel audio training objective: contrastive loss with a language model embedding as a teacher. Using over 500k speech segments from mental health audio interviews, we evaluate the utility of aligning Whisper's latent space with semantic representations from a text autoencoder (SBERT) and lexically derived embeddings of basic psychological dimensions: emotion and personality. Over self-supervised affective tasks and downstream psychological tasks, WhiSPA surpasses current speech encoders, achieving an average error reduction of 73.4% and 83.8%, respectively. WhiSPA demonstrates that it is not always necessary to run a subsequent text LM on speech-to-text output in order to get a rich psychological representation of human communication.
△ Less
Submitted 20 May, 2025; v1 submitted 15 January, 2025;
originally announced January 2025.
-
Unifying the Extremes: Developing a Unified Model for Detecting and Predicting Extremist Traits and Radicalization
Authors:
Allison Lahnala,
Vasudha Varadarajan,
Lucie Flek,
H. Andrew Schwartz,
Ryan L. Boyd
Abstract:
The proliferation of ideological movements into extremist factions via social media has become a global concern. While radicalization has been studied extensively within the context of specific ideologies, our ability to accurately characterize extremism in more generalizable terms remains underdeveloped. In this paper, we propose a novel method for extracting and analyzing extremist discourse acr…
▽ More
The proliferation of ideological movements into extremist factions via social media has become a global concern. While radicalization has been studied extensively within the context of specific ideologies, our ability to accurately characterize extremism in more generalizable terms remains underdeveloped. In this paper, we propose a novel method for extracting and analyzing extremist discourse across a range of online community forums. By focusing on verbal behavioral signatures of extremist traits, we develop a framework for quantifying extremism at both user and community levels. Our research identifies 11 distinct factors, which we term ``The Extremist Eleven,'' as a generalized psychosocial model of extremism. Applying our method to various online communities, we demonstrate an ability to characterize ideologically diverse communities across the 11 extremist traits. We demonstrate the power of this method by analyzing user histories from members of the incel community. We find that our framework accurately predicts which users join the incel community up to 10 months before their actual entry with an AUC of $>0.6$, steadily increasing to AUC ~0.9 three to four months before the event. Further, we find that upon entry into an extremist forum, the users tend to maintain their level of extremism within the community, while still remaining distinguishable from the general online discourse. Our findings contribute to the study of extremism by introducing a more holistic, cross-ideological approach that transcends traditional, trait-specific models.
△ Less
Submitted 8 January, 2025;
originally announced January 2025.
-
PsychAdapter: Adapting LLM Transformers to Reflect Traits, Personality and Mental Health
Authors:
Huy Vu,
Huy Anh Nguyen,
Adithya V Ganesan,
Swanie Juhng,
Oscar N. E. Kjell,
Joao Sedoc,
Margaret L. Kern,
Ryan L. Boyd,
Lyle Ungar,
H. Andrew Schwartz,
Johannes C. Eichstaedt
Abstract:
Artificial intelligence-based language generators are now a part of most people's lives. However, by default, they tend to generate "average" language without reflecting the ways in which people differ. Here, we propose a lightweight modification to the standard language model transformer architecture - "PsychAdapter" - that uses empirically derived trait-language patterns to generate natural lang…
▽ More
Artificial intelligence-based language generators are now a part of most people's lives. However, by default, they tend to generate "average" language without reflecting the ways in which people differ. Here, we propose a lightweight modification to the standard language model transformer architecture - "PsychAdapter" - that uses empirically derived trait-language patterns to generate natural language for specified personality, demographic, and mental health characteristics (with or without prompting). We applied PsychAdapters to modify OpenAI's GPT-2, Google's Gemma, and Meta's Llama 3 and found generated text to reflect the desired traits. For example, expert raters evaluated PsychAdapter's generated text output and found it matched intended trait levels with 87.3% average accuracy for Big Five personalities, and 96.7% for depression and life satisfaction. PsychAdapter is a novel method to introduce psychological behavior patterns into language models at the foundation level, independent of prompting, by influencing every transformer layer. This approach can create chatbots with specific personality profiles, clinical training tools that mirror language associated with psychological conditionals, and machine translations that match an authors reading or education level without taking up LLM context windows. PsychAdapter also allows for the exploration psychological constructs through natural language expression, extending the natural language processing toolkit to study human psychology.
△ Less
Submitted 31 December, 2024; v1 submitted 22 December, 2024;
originally announced December 2024.
-
Explaining GPT-4's Schema of Depression Using Machine Behavior Analysis
Authors:
Adithya V Ganesan,
Vasudha Varadarajan,
Yash Kumar Lal,
Veerle C. Eijsbroek,
Katarina Kjell,
Oscar N. E. Kjell,
Tanuja Dhanasekaran,
Elizabeth C. Stade,
Johannes C. Eichstaedt,
Ryan L. Boyd,
H. Andrew Schwartz,
Lucie Flek
Abstract:
Use of large language models such as ChatGPT (GPT-4) for mental health support has grown rapidly, emerging as a promising route to assess and help people with mood disorders, like depression. However, we have a limited understanding of GPT-4's schema of mental disorders, that is, how it internally associates and interprets symptoms. In this work, we leveraged contemporary measurement theory to dec…
▽ More
Use of large language models such as ChatGPT (GPT-4) for mental health support has grown rapidly, emerging as a promising route to assess and help people with mood disorders, like depression. However, we have a limited understanding of GPT-4's schema of mental disorders, that is, how it internally associates and interprets symptoms. In this work, we leveraged contemporary measurement theory to decode how GPT-4 interrelates depressive symptoms to inform both clinical utility and theoretical understanding. We found GPT-4's assessment of depression: (a) had high overall convergent validity (r = .71 with self-report on 955 samples, and r = .81 with experts judgments on 209 samples); (b) had moderately high internal consistency (symptom inter-correlates r = .23 to .78 ) that largely aligned with literature and self-report; except that GPT-4 (c) underemphasized suicidality's -- and overemphasized psychomotor's -- relationship with other symptoms, and (d) had symptom inference patterns that suggest nuanced hypotheses (e.g. sleep and fatigue are influenced by most other symptoms while feelings of worthlessness/guilt is mostly influenced by depressed mood).
△ Less
Submitted 20 November, 2024;
originally announced November 2024.
-
Towards a Virtual Reality Visualization of Hand-Object Interactions to Support Remote Physical Therapy
Authors:
Trudi Di Qi,
LouAnne Boyd,
Scott Fitzpatrick,
Meghna Raswan,
Farnceli Cibrian
Abstract:
Improving object manipulation skills through hand-object interaction exercises is crucial for rehabilitation. Despite limited healthcare resources, physical therapists propose remote exercise routines followed up by remote monitoring. However, remote motor skills assessment remains challenging due to the lack of effective motion visualizations. Therefore, exploring innovative ways of visualization…
▽ More
Improving object manipulation skills through hand-object interaction exercises is crucial for rehabilitation. Despite limited healthcare resources, physical therapists propose remote exercise routines followed up by remote monitoring. However, remote motor skills assessment remains challenging due to the lack of effective motion visualizations. Therefore, exploring innovative ways of visualization is crucial, and virtual reality (VR) has shown the potential to address this limitation. However, it is unclear how VR visualization can represent understandable hand-object interactions. To address this gap, in this paper, we present VRMoVi, a VR visualization system that incorporates multiple levels of 3D visualization layers to depict movements. In a 2-stage study, we showed VRMoVi's potential in representing hand-object interactions, with its visualization outperforming traditional representations, and detailed features improved the hand-object interactions understanding. This study takes the initial step in developing VR visualization of hand-object interaction to support remote physical therapy.
△ Less
Submitted 11 December, 2023; v1 submitted 22 March, 2023;
originally announced March 2023.
-
Learning Physics Properties of Fabrics and Garments with a Physics Similarity Neural Network
Authors:
Li Duan,
Lewis Boyd,
Gerardo Aragon-Camarasa
Abstract:
In this paper, we propose to predict the physics parameters of real fabrics and garments by learning their physics similarities between simulated fabrics via a Physics Similarity Network (PhySNet). For this, we estimate wind speeds generated by an electric fan and the area weight to predict bending stiffness of simulated and real fabrics and garments. We found that PhySNet coupled with a Bayesian…
▽ More
In this paper, we propose to predict the physics parameters of real fabrics and garments by learning their physics similarities between simulated fabrics via a Physics Similarity Network (PhySNet). For this, we estimate wind speeds generated by an electric fan and the area weight to predict bending stiffness of simulated and real fabrics and garments. We found that PhySNet coupled with a Bayesian optimiser can predict physics parameters and improve the state-of-art by 34%for real fabrics and 68% for real garments.
△ Less
Submitted 20 December, 2021;
originally announced December 2021.
-
Simulation of the Response of the Solid State Neutron Detector for the European Spallation Source
Authors:
L. Boyd,
E. Rofors,
J. R. M. Annand,
K. G. Fissum,
R. Hall-Wilton,
R. Al Jebali,
K. Kanaki,
K. Livingston,
V. Maulerova,
N. Mauritzson,
R. Montgomery,
H. Perrey,
B. Seitz
Abstract:
The characteristics of the Solid-state Neutron Detector, under development for neutron-scattering measurements at the European Spallation Source, have been simulated with a Geant4-based computer code. The code models the interations of thermal neutrons and ionising radiation in the 6Li-doped scintillating glass of the detector, the production of scintillation light and the transport of optical, sc…
▽ More
The characteristics of the Solid-state Neutron Detector, under development for neutron-scattering measurements at the European Spallation Source, have been simulated with a Geant4-based computer code. The code models the interations of thermal neutrons and ionising radiation in the 6Li-doped scintillating glass of the detector, the production of scintillation light and the transport of optical, scintillation photons through the the scintillator, en route to the photo-cathode of the attached multi-anode photomultiplier. Factors which affect the optical-photon transport, such as surface finish, pixelation of the glass sheet, provision of a front reflector and optical coupling media are compared. Predictions of the detector response are compared with measurements made with neutron and gamma-ray sources, a collimated alpha source and finely collimated beams of 2.5 MeV protons and deuterons.
△ Less
Submitted 11 March, 2021; v1 submitted 29 October, 2020;
originally announced October 2020.
-
Response of a Li-glass/multi-anode photomultiplier detector to collimated thermal-neutron beams
Authors:
E. Rofors,
N. Mauritzson,
H. Perrey,
R. Al Jebali,
J. R. M. Annand,
L. Boyd,
M. J. Christensen,
U. Clemens,
S. Desert,
R. Engels,
K. G. Fissum,
H. Frielinghaus,
C. Gheorghe,
R. Hall-Wilton,
S. Jaksch,
K. Kanaki,
S. Kazi,
G. Kemmerling,
I. Llamas Jansa,
V. Maulerova,
R. Montgomery,
T. Richter,
J. Scherzinger,
B. Seitz,
M. Shetty
Abstract:
The response of a position-sensitive Li-glass scintillator detector being developed for thermal-neutron detection with 6 mm position resolution has been investigated using collimated beams of thermal neutrons. The detector was moved perpendicularly through the neutron beams in 0.5 to 1.0 mm horizontal and vertical steps. Scintillation was detected in an 8 X 8 pixel multi-anode photomultiplier tube…
▽ More
The response of a position-sensitive Li-glass scintillator detector being developed for thermal-neutron detection with 6 mm position resolution has been investigated using collimated beams of thermal neutrons. The detector was moved perpendicularly through the neutron beams in 0.5 to 1.0 mm horizontal and vertical steps. Scintillation was detected in an 8 X 8 pixel multi-anode photomultiplier tube on an event-by-event basis. In general, several pixels registered large signals at each neutron-beam location. The number of pixels registering signal above a set threshold was investigated, with the maximization of the single-hit efficiency over the largest possible area of the detector as the primary goal. At a threshold of ~50% of the mean of the full-deposition peak, ~80% of the events were registered in a single pixel, resulting in an effective position resolution of ~5 mm in X and Y. Lower thresholds generally resulted in events demonstrating higher pixel multiplicities, but these events could also be localized with ~5 mm position resolution.
△ Less
Submitted 9 December, 2020; v1 submitted 13 October, 2020;
originally announced October 2020.
-
Response of a Li-glass/multi-anode photomultiplier detector to focused proton and deuteron beams
Authors:
E. Rofors,
J. Pallon,
R. Al Jebali,
J. R. M. Annand,
L. Boyd,
M. J. Christensen,
U. Clemens,
S. Desert,
M. Elfman,
R. Engels,
K. G. Fissum,
H. Frielinghaus,
R. Frost,
S. Gardner,
C. Gheorghe,
R. Hall-Wilton,
S. Jaksch,
K. Kanaki,
G. Kemmerling,
P. Kristiansson,
K. Livingston,
V. Maulerova,
N. Mauritzson,
R. Montgomery,
H. Perrey
, et al. (4 additional authors not shown)
Abstract:
The response of a position-sensitive Li-glass based scintillation detector to focused beams of 2.5 MeV protons and deuterons has been investigated. The beams were scanned across the detector in 0.5 mm horizontal and vertical steps perpendicular to the beams. Scintillation light was registered using an 8 by 8 pixel multi-anode photomultiplier tube. The signal amplitudes were recorded for each pixel…
▽ More
The response of a position-sensitive Li-glass based scintillation detector to focused beams of 2.5 MeV protons and deuterons has been investigated. The beams were scanned across the detector in 0.5 mm horizontal and vertical steps perpendicular to the beams. Scintillation light was registered using an 8 by 8 pixel multi-anode photomultiplier tube. The signal amplitudes were recorded for each pixel on an event-by-event basis. Several pixels generally registered considerable signals at each beam location. The number of pixels above set thresholds were investigated, with the optimization of the single-hit efficiency over the largest possible area as the goal. For both beams, at a threshold of ~50% of the mean of the full-deposition peak, ~80% of the events were registered in a single pixel, resulting in an effective position resolution of ~5 mm in X and Y.
△ Less
Submitted 18 May, 2020;
originally announced May 2020.
-
Response of a Li-glass/multi-anode photomultiplier detector to $α$-particles from $^{241}$Am
Authors:
E. Rofors,
H. Perrey,
R. Al Jebali,
J. R. M. Annand,
L. Boyd,
U. Clemens,
S. Desert,
R. Engels,
K. G. Fissum,
H. Frielinghaus,
C. Gheorghe,
R. Hall-Wilton,
S. Jaksch,
A. Jalgén,
K. Kanaki,
G. Kemmerling,
V. Maulerova,
N. Mauritzson,
R. Montgomery,
J. Scherzinger,
B. Seitz
Abstract:
The response of a position-sensitive Li-glass scintillator detector to $α$-particles from a collimated $^{241}$Am source scanned across the face of the detector has been measured. Scintillation light was read out by an 8 X 8 pixel multi-anode photomultiplier and the signal amplitude for each pixel has been recorded for every position on a scan. The pixel signal is strongly dependent on position an…
▽ More
The response of a position-sensitive Li-glass scintillator detector to $α$-particles from a collimated $^{241}$Am source scanned across the face of the detector has been measured. Scintillation light was read out by an 8 X 8 pixel multi-anode photomultiplier and the signal amplitude for each pixel has been recorded for every position on a scan. The pixel signal is strongly dependent on position and in general several pixels will register a signal (a hit) above a given threshold. The effect of this threshold on hit multiplicity is studied, with a view to optimize the single-hit efficiency of the detector.
△ Less
Submitted 19 December, 2018;
originally announced December 2018.
-
The neutron tagging facility at Lund University
Authors:
F. Messi,
H. Perrey,
K. Fissum,
M. Akkawi,
R. Al Jebali,
J. R. M. Annand,
P. Bentley,
L. Boyd,
C. P. Cooper-Jensen,
D. D. DiJulio,
J. Freita-Ramos,
R. Hall-Wilton,
A. Huusko,
T. Ilves,
F. Issa,
A. Jalgén,
K. Kanaki,
E. Karnickis,
A. Khaplanov,
S. Koufigar,
V. Maulerova,
G. Mauri,
N. Mauritzson,
W. Pei,
F. Piscitelli
, et al. (5 additional authors not shown)
Abstract:
Over the last decades, the field of thermal neutron detection has overwhelmingly employed He-3-based technologies. The He-3 crisis together with the forthcoming establishment of the European Spallation Source have necessitated the development of new technologies for neutron detection. Today, several promising He-3-free candidates are under detailed study and need to be validated. This validation p…
▽ More
Over the last decades, the field of thermal neutron detection has overwhelmingly employed He-3-based technologies. The He-3 crisis together with the forthcoming establishment of the European Spallation Source have necessitated the development of new technologies for neutron detection. Today, several promising He-3-free candidates are under detailed study and need to be validated. This validation process is in general long and expensive. The study of detector prototypes using neutron-emitting radioactive sources is a cost-effective solution, especially for preliminary investigations. That said, neutron-emitting sources have the general disadvantage of broad, structured, emitted-neutron energy ranges. Further, the emitted neutrons often compete with unwanted backgrounds of gamma-rays, alpha-particles, and fission-fragments. By blending experimental infrastructure such as shielding to provide particle beams with neutron-detection techniques such as tagging, disadvantages may be converted into advantages. In particular, a technique known as tagging involves exploiting the mixed-field generally associated with a neutron-emitting source to determine neutron time-of-flight and thus energy on an event-by-event basis. This allows for the definition of low-cost, precision neutron beams. The Source-Testing Facility, located at Lund University in Sweden and operated by the SONNIG Group of the Division of Nuclear Physics, was developed for just such low-cost studies. Precision tagged-neutron beams derived from radioactive sources are available around-the-clock for advanced detector diagnostic studies. Neutron measurements performed at the Source Testing Facility are thus cost-effective and have a very low barrier for entry. In this paper, we present an overview of the project.
△ Less
Submitted 28 November, 2017;
originally announced November 2017.
-
The N2K Consortium. II. A Transiting Hot Saturn Around HD 149026 With a Large Dense Core
Authors:
B. Sato,
D. A. Fischer,
G. W. Henry,
G. Laughlin,
R. P. Butler,
G. W. Marcy,
S. S. Vogt,
P. Bodenheimer,
S. Ida,
E. Toyota,
A. Wolf,
J. A. Valenti,
L. J. Boyd,
J. A. Johnson,
J. T. Wright,
M. Ammons,
S. Robinson,
J. Strader,
C. McCarthy,
K. L. Tah,
D. Minniti
Abstract:
Doppler measurements from Subaru and Keck have revealed radial velocity variations in the V=8.15, G0IV star HD 149026 consistent with a Saturn-Mass planet in a 2.8766 day orbit. Photometric observations at Fairborn Observatory have detected three complete transit events with depths of 0.003 mag at the predicted times of conjunction. HD 149026 is now the second brightest star with a transiting ex…
▽ More
Doppler measurements from Subaru and Keck have revealed radial velocity variations in the V=8.15, G0IV star HD 149026 consistent with a Saturn-Mass planet in a 2.8766 day orbit. Photometric observations at Fairborn Observatory have detected three complete transit events with depths of 0.003 mag at the predicted times of conjunction. HD 149026 is now the second brightest star with a transiting extrasolar planet. The mass of the star, based on interpolation of stellar evolutionary models, is 1.3 +/- 0.1 solar masses; together with the Doppler amplitude, K=43.3 m s^-1, we derive a planet mass Msin(i)=0.36 Mjup, and orbital radius of 0.042 AU. HD 149026 is chromospherically inactive and metal-rich with spectroscopically derived [Fe/H]=+0.36, Teff=6147 K, log g=4.26 and vsin(i)=6.0 km s^-1. Based on Teff and the stellar luminosity of 2.72 Lsun, we derive a stellar radius of 1.45 Rsun. Modeling of the three photometric transits provides an orbital inclination of 85.3 +/- 1.0 degrees and (including the uncertainty in the stellar radius) a planet radius of 0.725 +/- 0.05 Rjup. Models for this planet mass and radius suggest the presence of a ~67 Mearth core composed of elements heavier than hydrogen and helium. This substantial planet core would be difficult to construct by gravitational instability.
△ Less
Submitted 30 June, 2005;
originally announced July 2005.
-
The ASTRA Spectrophotometer: A July 2004 Progress Report
Authors:
S. J. Adelman,
A. F. Gulliver,
B. Smalley,
J. S. Pazder,
P. F. Younger,
L. Boyd,
D. Epand
Abstract:
A cross-dispersed spectrophotometer with CCD detector and its automated 0.5-m telescope at the Fairborn Observatory, now under construction, should begin observations in 6 to 9 months. The Citadel ASTRA Telescope will be able to observe Vega the primary standard, make rapid measurements of the naked-eye stars, use 10 minutes per hour to obtain photometric measurements of the nightly extinction,…
▽ More
A cross-dispersed spectrophotometer with CCD detector and its automated 0.5-m telescope at the Fairborn Observatory, now under construction, should begin observations in 6 to 9 months. The Citadel ASTRA Telescope will be able to observe Vega the primary standard, make rapid measurements of the naked-eye stars, use 10 minutes per hour to obtain photometric measurements of the nightly extinction, and obtain high quality observations of V=10.5 mag. stars in an hour. The approximate wavelength range is 3300-9000A with a resolution of 14A in first and 7A in second order. Filter photometric magnitudes and indices will be calibrated in part for use as quality checks.
Science observations for major projects such as comparisons with model atmospheres codes and for exploratory investigations should also begin in the first year. The ASTRA team realizes to deal with this potential data flood that they will need help to make the best scientific uses of the data. Thus they are interested in discussing possible collaborations. In less than a year of normal observing, all isolated stars in the Bright Star Catalog that can be observed can have their fluxes well measured. Some A Star related applications are discussed.
△ Less
Submitted 6 September, 2004;
originally announced September 2004.
