-
Emergence of psychopathological computations in large language models
Authors:
Soo Yong Lee,
Hyunjin Hwang,
Taekwan Kim,
Yuyeong Kim,
Kyuri Park,
Jaemin Yoo,
Denny Borsboom,
Kijung Shin
Abstract:
Can large language models (LLMs) implement computations of psychopathology? An effective approach to the question hinges on addressing two factors. First, for conceptual validity, we require a general and computational account of psychopathology that is applicable to computational entities without biological embodiment or subjective experience. Second, mechanisms underlying LLM behaviors need to b…
▽ More
Can large language models (LLMs) implement computations of psychopathology? An effective approach to the question hinges on addressing two factors. First, for conceptual validity, we require a general and computational account of psychopathology that is applicable to computational entities without biological embodiment or subjective experience. Second, mechanisms underlying LLM behaviors need to be studied for better methodological validity. Thus, we establish a computational-theoretical framework to provide an account of psychopathology applicable to LLMs. To ground the theory for empirical analysis, we also propose a novel mechanistic interpretability method alongside a tailored empirical analytic framework. Based on the frameworks, we conduct experiments demonstrating three key claims: first, that distinct dysfunctional and problematic representational states are implemented in LLMs; second, that their activations can spread and self-sustain to trap LLMs; and third, that dynamic, cyclic structural causal models encoded in the LLMs underpin these patterns. In concert, the empirical results corroborate our hypothesis that network-theoretic computations of psychopathology have already emerged in LLMs. This suggests that certain LLM behaviors mirroring psychopathology may not be a superficial mimicry but a feature of their internal processing. Thus, our work alludes to the possibility of AI systems with psychopathological behaviors in the near future.
△ Less
Submitted 10 April, 2025;
originally announced April 2025.
-
Efficient Neural Network Approximation of Robust PCA for Automated Analysis of Calcium Imaging Data
Authors:
Seungjae Han,
Eun-Seo Cho,
Inkyu Park,
Kijung Shin,
Young-Gyu Yoon
Abstract:
Calcium imaging is an essential tool to study the activity of neuronal populations. However, the high level of background fluorescence in images hinders the accurate identification of neurons and the extraction of neuronal activities. While robust principal component analysis (RPCA) is a promising method that can decompose the foreground and background in such images, its computational complexity…
▽ More
Calcium imaging is an essential tool to study the activity of neuronal populations. However, the high level of background fluorescence in images hinders the accurate identification of neurons and the extraction of neuronal activities. While robust principal component analysis (RPCA) is a promising method that can decompose the foreground and background in such images, its computational complexity and memory requirement are prohibitively high to process large-scale calcium imaging data. Here, we propose BEAR, a simple bilinear neural network for the efficient approximation of RPCA which achieves an order of magnitude speed improvement with GPU acceleration compared to the conventional RPCA algorithms. In addition, we show that BEAR can perform foreground-background separation of calcium imaging data as large as tens of gigabytes. We also demonstrate that two BEARs can be cascaded to perform simultaneous RPCA and non-negative matrix factorization for the automated extraction of spatial and temporal footprints from calcium imaging data. The source code used in the paper is available at https://github.com/NICALab/BEAR.
△ Less
Submitted 30 July, 2021;
originally announced August 2021.
-
Theoretical investigation on the Schelling's critical neighborhood demand
Authors:
Jae Kyun Shin,
Hiroki Sayama
Abstract:
We derived the critical neighborhood demand in the Schelling's segregation model by studying the conditions for which a chain reaction of migrations of unsatisfied agents occurs. The essence of Schelling dynamics was approximated in two simplified models: (1) a random walk model for the initial stage of the migrations to illustrate the power-law behavior of chain reaction lengths under critical co…
▽ More
We derived the critical neighborhood demand in the Schelling's segregation model by studying the conditions for which a chain reaction of migrations of unsatisfied agents occurs. The essence of Schelling dynamics was approximated in two simplified models: (1) a random walk model for the initial stage of the migrations to illustrate the power-law behavior of chain reaction lengths under critical conditions, and (2) a two-room model for the whole process to represent a non-spatial version of segregation dynamics in the Schelling model. Our theoretical results showed good agreements with numerical results obtained from agent-based simulations.
△ Less
Submitted 5 February, 2014;
originally announced February 2014.
-
A state equation for the Schelling's segregation model
Authors:
Jae Kyun Shin,
Hiroki Sayama,
Seung Ryul Choi
Abstract:
An aspatial version for the famous Schelling's segregation model has recently been proposed, which, called two-room model, is still in an agent-based format like the original Schelling model. In the present study, we propose a new, state equation version of the Schelling model. The new equation is based on the two-room model and is derived in terms of a set of discrete maps. Fixed point solutions…
▽ More
An aspatial version for the famous Schelling's segregation model has recently been proposed, which, called two-room model, is still in an agent-based format like the original Schelling model. In the present study, we propose a new, state equation version of the Schelling model. The new equation is based on the two-room model and is derived in terms of a set of discrete maps. Fixed point solutions for the new equation are found analytically and confirmed numerically. Especially, we show that the extremely simple state equations can reasonably reveal the essence of the Schelling dynamics: integration, segregation and tipping. In addition to the fixed point solutions, periodic solutions are identified and conditions of the limit cycles are derived analytically.
△ Less
Submitted 5 February, 2014;
originally announced February 2014.
-
Tipping Diffusivity in Information Accumulation Systems: More Links, Less Consensus
Authors:
Jae Kyun Shin,
Jan Lorenz
Abstract:
Assume two different communities each of which maintain their respective opinions mainly because of the weak interaction between the two communities. In such a case, it is an interesting problem to find the necessary strength of inter-community interaction in order for the two communities to reach a consensus. In this paper, the Information Accumulation System (IAS) model is applied to investigate…
▽ More
Assume two different communities each of which maintain their respective opinions mainly because of the weak interaction between the two communities. In such a case, it is an interesting problem to find the necessary strength of inter-community interaction in order for the two communities to reach a consensus. In this paper, the Information Accumulation System (IAS) model is applied to investigate the problem. With the application of the IAS model, the opinion dynamics of the two-community problem is found to belong to a wider class of two-species problems appearing in population dynamics or in competition of two languages, for all of which the governing equations can be described in terms of coupled logistic maps. Tipping diffusivity is defined as the maximal inter-community interaction that the two communities can stay in different opinions. For a problem with simple community structure and homogeneous individuals, the tipping diffusivity is calculated theoretically. As a conclusion of the paper, the convergence of the two communities to the same value is less possible the more overall interaction, intra-community and inter-community, takes place. This implies, for example, that the increase in the interaction between individuals caused by the development of modern communication tools, such as facebook and twitter, does not necessarily improve the tendency towards global convergence between different communities. If the number of internal links increases by a factor, the number of inter-community links must be increased by an even higher factor, in order for consensus to be the only stable attractor.
△ Less
Submitted 9 May, 2010; v1 submitted 7 September, 2009;
originally announced September 2009.
