pith. sign in

arxiv: 2605.20363 · v1 · pith:2WGSFRMTnew · submitted 2026-05-19 · 💻 cs.SI

Mapping the Winds of Stance Dynamics using Potential Landscape Models

Benjamin Steel , Derek Ruths This is my paper

Pith reviewed 2026-05-21 07:13 UTC · model grok-4.3

classification 💻 cs.SI
keywords stance detectionpotential landscapepublic opiniondimensionality reductionsocial medialatent spacepolitical shiftsdynamics modeling
0
0 comments X

The pith

A method using stance detection, dimensionality reduction, and neural networks recovers a three-dimensional potential landscape explaining 45% of variance in public stance expressions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a pipeline that first extracts stance expressions from large social media datasets, reduces them to a low-dimensional linear space, and then trains neural networks to model the potential landscape whose gradients indicate directions of collective stance change. This setup is intended to capture multi-issue, population-level drifts that single-issue tracking methods miss, such as coordinated shifts by political figures over time and platforms. A reader would care if the approach works because it turns raw opinion data into an interpretable map of how groups move between positions on many topics at once. The authors demonstrate the pipeline on Canadian political figures and report that the resulting three-dimensional space explains 45% of the variance while allowing each dimension to be characterized.

Core claim

The authors show that large-scale stance expressions can be projected into a coherent linear three-dimensional latent space whose dynamics are approximated by a potential landscape recovered with neural networks; the landscape gradients then reveal en-mass stance shifts by prominent Canadian political figures across multiple platforms and years, and the space itself explains 45% of the variance in the original stance data with interpretable characteristics for each dimension.

What carries the argument

The potential landscape of stance dynamics, recovered by training neural networks on a low-dimensional linear latent space obtained after stance detection and dimensionality reduction.

If this is right

  • The gradient of the recovered potential landscape directly indicates the direction and strength of large-scale stance shifts at the population level.
  • The framework tracks opinion change across a diverse range of issues simultaneously rather than requiring advance selection of a single group and issue.
  • Application to real data from political figures produces a three-dimensional space whose axes have specific, describable characteristics.
  • The descriptive value of the landscape is supported even when predictive performance on future shifts is only mixed.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the low-dimensional landscape structure holds across other domains, the same pipeline could map shifts in fashion trends or economic policy views without new labeled data.
  • Treating stance movement as flow on a potential surface opens the possibility of borrowing stability or tipping-point analysis from physics to forecast sudden collective changes.
  • Testing whether the same three-dimensional structure appears in non-English or non-political corpora would show how general the reported variance capture is.
  • The mixed predictive results suggest that relaxing the linearity assumption in the latent space could improve forecasting while preserving the descriptive landscape view.

Load-bearing premise

Stance expressions are assumed to lie on a low-dimensional linear latent space whose dynamics can be well approximated by a potential landscape recoverable via neural networks.

What would settle it

Running the full pipeline of stance detection, dimensionality reduction, and potential-landscape neural networks on a fresh collection of social media posts about political figures and finding no three-dimensional space that explains approximately 45% of variance with clearly interpretable axes would falsify the central claim.

Figures

Figures reproduced from arXiv: 2605.20363 by Benjamin Steel, Derek Ruths.

Figure 1
Figure 1. Figure 1: ‘This year, people on the left have diverged in their stated preference for authoritarian or libertarian systems, but people of the right have converged in support for author￾itarianism’ —how can we re-construct this fictional, ideal￾ized now-cast in a data-driven, empirical fashion? deliberation, or high dimensional longitudinal survey data. We infer latent stance trajectories from a cross-platform datase… view at source ↗
Figure 2
Figure 2. Figure 2: Pipeline for inferring the dynamical stance landscape. (1) Stance targets are automatically extracted from the corpus. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Potential landscape predictive-ness on held-out [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: The first two dimensions of the latent stance space. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 6
Figure 6. Figure 6: Snapshots in time of the evolving stance density landscape and potential landscape, for the first two principal compo [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of the potential landscape of the 4 platforms in the dataset. We do dimensionality reduction on all account [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
read the original abstract

From changing fashion trends to views on world leaders and economic policies, large-scale shifts in group positions happen regularly and unexpectedly. How can we track these in the wild? How can we characterize them? Existing work has primarily leveraged stance detection to track shifts of specific groups on a single issue. However, such methods will only find shifts when they accurately pick exactly the right group and right issue. They do not capture the multi-dimensional, multi-resolution stance landscape in which these shifts actually happen. To better model drift and shift in public opinion, we require a framework that can track change at the population level, across a diverse range of issues. We propose a method to infer the potential landscape of stance dynamics, the gradient of which shows large-scale stance shifts, and apply it to show en-mass stance shifts by prominent Canadian political figures across multiple platforms and years. We do this using large-scale stance detection to find stance expressions, dimensionality reduction to find the low-dimensional linear latent space, and potential landscape neural networks to find the potential landscape of that space. This allows us to find a coherent, linear, three-dimensional space that explains 45\% of the variance in stance, where we can explain the specific characteristics of each dimension. We show that while the predictive performance is sufficient to validate its descriptive-ness, in practice its predictive performance is mixed.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes a framework to model multi-dimensional stance dynamics in public opinion by combining large-scale stance detection, linear dimensionality reduction to recover a three-dimensional latent space explaining 45% of variance, and neural networks to infer a potential landscape whose gradients represent large-scale stance shifts ('winds'). The method is applied to stance expressions from prominent Canadian political figures across platforms and years, with the claim that the resulting dimensions are coherent and interpretable while the overall approach has sufficient descriptive power despite mixed predictive performance.

Significance. If the modeling assumptions hold, the work could offer a useful population-level view of stance change that goes beyond single-issue tracking. The combination of linear embedding and potential-landscape recovery is a novel direction for computational social science. However, the modest variance explained and the mixed predictive results limit the immediate strength of the conclusions about recovered dynamics.

major comments (3)
  1. [Abstract] Abstract: the claim of a 'coherent, linear, three-dimensional space that explains 45% of the variance in stance' is central to the contribution, yet the same paragraph reports mixed predictive performance; this raises the question whether the linear embedding and subsequent neural-network potential recovery actually capture real gradient flows or are shaped by the fitting choices on the same data.
  2. [Method] Method (pipeline description): the assumption that stance dynamics lie on a low-dimensional linear latent space whose vector field is the gradient of a scalar potential recoverable by neural networks is load-bearing; no ablation against a general (non-conservative) vector-field model or against a non-linear embedding is described, leaving open whether the three explainable dimensions and inferred winds are data properties or modeling artifacts.
  3. [Results] Results (application to Canadian political figures): the reported en-mass stance shifts rest on the recovered landscape; without external benchmarks or comparison to higher-dimensional baselines, the 45% variance figure and dimension interpretations risk circularity from fitting both the dimensionality reduction and the potential network to the identical stance data.
minor comments (2)
  1. [Abstract] The phrase 'sufficient to validate its descriptive-ness' in the abstract would benefit from a quantitative definition or cross-validation metric to clarify what threshold is being used.
  2. [Figures] Figure captions for the landscape visualizations should specify the meaning of axes, color scales, and any normalization applied to the potential or gradient arrows.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback on our manuscript. We address each major comment point by point below, clarifying our methodological choices and indicating revisions made to improve clarity and address concerns about potential artifacts or circularity.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of a 'coherent, linear, three-dimensional space that explains 45% of the variance in stance' is central to the contribution, yet the same paragraph reports mixed predictive performance; this raises the question whether the linear embedding and subsequent neural-network potential recovery actually capture real gradient flows or are shaped by the fitting choices on the same data.

    Authors: The 45% variance figure is obtained exclusively from the linear dimensionality reduction (PCA) step performed on the high-dimensional stance vectors, prior to and independent of the neural potential landscape fitting. The potential network is subsequently trained on the observed trajectories within this reduced space to recover a scalar function whose negative gradient approximates the empirical shifts. We acknowledge that mixed predictive performance indicates limitations in forecasting unseen data, which we report transparently; however, the primary contribution is descriptive visualization of population-level dynamics rather than strong predictive accuracy. We have revised the abstract to explicitly separate the embedding variance from the potential recovery step and to qualify that the inferred winds represent gradients fitted to the observed flows in the latent space. revision: yes

  2. Referee: [Method] Method (pipeline description): the assumption that stance dynamics lie on a low-dimensional linear latent space whose vector field is the gradient of a scalar potential recoverable by neural networks is load-bearing; no ablation against a general (non-conservative) vector-field model or against a non-linear embedding is described, leaving open whether the three explainable dimensions and inferred winds are data properties or modeling artifacts.

    Authors: We selected the linear embedding specifically to preserve interpretability, allowing each dimension to be characterized by its loadings on specific stance issues, which aligns with the goal of explaining the characteristics of the recovered space. The conservative potential assumption follows standard modeling practices in dynamical systems for opinion and social change, where a scalar potential is used to represent driving forces. While we did not perform explicit ablations against non-linear embeddings or non-conservative vector fields, we have added a dedicated paragraph in the Methods and a limitations subsection in the Discussion to justify these choices on grounds of interpretability and computational feasibility, and to note that alternative models remain an important direction for future work. revision: partial

  3. Referee: [Results] Results (application to Canadian political figures): the reported en-mass stance shifts rest on the recovered landscape; without external benchmarks or comparison to higher-dimensional baselines, the 45% variance figure and dimension interpretations risk circularity from fitting both the dimensionality reduction and the potential network to the identical stance data.

    Authors: The dimensionality reduction is performed unsupervised on the full set of stance vectors, after which trajectories are projected and the potential is fitted to the resulting time-series displacements; the dimension interpretations are derived from the highest-loading stance issues on each principal component, which correspond to recognizable political axes. To address potential circularity, we have added a supplementary analysis comparing variance explained against randomized baselines and higher-dimensional projections, and we include qualitative alignment checks with documented political events. While domain-specific external benchmarks for multi-issue stance shifts are not readily available, these additions reduce reliance on the fitted landscape alone for validating the observed shifts. revision: yes

Circularity Check

0 steps flagged

No significant circularity: results follow directly from applying dimensionality reduction and NN fitting to stance data

full rationale

The paper describes a pipeline of stance detection, dimensionality reduction to a linear latent space, and neural network recovery of a potential landscape on that space. The 45% variance explained and dimension characteristics are direct outputs of the fitted models rather than predictions or derivations that reduce to the inputs by construction. No self-citations, uniqueness theorems, or ansatzes smuggled via prior work are referenced in the abstract or described steps. The explicit note of mixed predictive performance further indicates the authors present the landscape as a descriptive fit, not a forced or self-referential result. The chain is a standard modeling application and remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the existence of a recoverable low-dimensional linear latent space and on the validity of modeling its dynamics as a potential landscape; both are domain assumptions rather than derived results.

free parameters (1)
  • number of latent dimensions
    Set to three to achieve 45% explained variance; chosen after reduction rather than fixed a priori.
axioms (1)
  • domain assumption Stance expressions admit a low-dimensional linear embedding whose dynamics follow a potential landscape
    Invoked by the sequence of dimensionality reduction followed by potential landscape neural networks.

pith-pipeline@v0.9.0 · 5762 in / 1311 out tokens · 36558 ms · 2026-05-21T07:13:27.209457+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

97 extracted references · 97 canonical work pages · 2 internal anchors

  1. [1]

    Smooth regression analysis , author=. Sankhy. 1964 , publisher=

    work page 1964

  2. [2]

    Proceedings of the International AAAI Conference on Web and Social Media , volume=

    Measuring belief dynamics on twitter , author=. Proceedings of the International AAAI Conference on Web and Social Media , volume=

    work page

  3. [3]

    Nature Human Behaviour , pages=

    A semantic embedding space based on large language models for modelling human beliefs , author=. Nature Human Behaviour , pages=. 2025 , publisher=

    work page 2025

  4. [4]

    The Journal of physiology , volume=

    Bioattractors: dynamical systems theory and the evolution of regulatory processes , author=. The Journal of physiology , volume=. 2014 , publisher=

    work page 2014

  5. [5]

    Proceedings of the National Academy of Sciences , volume=

    Quantifying the Waddington landscape and biological paths for development and differentiation , author=. Proceedings of the National Academy of Sciences , volume=. 2011 , publisher=

    work page 2011

  6. [6]

    bioRxiv , pages=

    Reconstructing waddington’s landscape from data , author=. bioRxiv , pages=. 2025 , publisher=

    work page 2025

  7. [7]

    Nature Communications , volume=

    Active droploids , author=. Nature Communications , volume=. 2021 , publisher=

    work page 2021

  8. [8]

    Reviews of modern physics , volume=

    Hydrodynamics of soft active matter , author=. Reviews of modern physics , volume=. 2013 , publisher=

    work page 2013

  9. [9]

    Annual review of fluid mechanics , volume=

    Lagrangian coherent structures , author=. Annual review of fluid mechanics , volume=. 2015 , publisher=

    work page 2015

  10. [10]

    PLOS Computational Biology , volume=

    The specious art of single-cell genomics , author=. PLOS Computational Biology , volume=. 2023 , publisher=

    work page 2023

  11. [11]

    2014 , publisher=

    The strategy of the genes , author=. 2014 , publisher=

    work page 2014

  12. [12]

    Physical Review X , volume=

    Learning geometric models for developmental dynamics , author=. Physical Review X , volume=. 2025 , publisher=

    work page 2025

  13. [13]

    bioRxiv , pages=

    Generative epigenetic landscapes map the topology and topography of cell fates , author=. bioRxiv , pages=. 2025 , publisher=

    work page 2025

  14. [14]

    2023 , publisher=

    Persuasion in parallel: How information changes minds about politics , author=. 2023 , publisher=

    work page 2023

  15. [15]

    international conference on machine learning , pages=

    Dropout as a bayesian approximation: Representing model uncertainty in deep learning , author=. international conference on machine learning , pages=. 2016 , organization=

    work page 2016

  16. [16]

    Proceedings of the 6th Workshop on Computational Approaches to Discourse, Context and Document-Level Inferences (CODI 2025) , pages=

    Corpus-Oriented Stance Target Extraction , author=. Proceedings of the 6th Workshop on Computational Approaches to Discourse, Context and Document-Level Inferences (CODI 2025) , pages=

    work page 2025

  17. [17]

    Bioinformatics , volume=

    Missing value estimation methods for DNA microarrays , author=. Bioinformatics , volume=. 2001 , publisher=

    work page 2001

  18. [18]

    Alex Rubinsteyn and Sergey Feldman , title=

    work page

  19. [19]

    Neural computation , volume=

    Slow feature analysis: Unsupervised learning of invariances , author=. Neural computation , volume=. 2002 , publisher=

    work page 2002

  20. [20]

    Journal of Nonlinear Science , volume=

    Variational approach for learning Markov processes from time series data , author=. Journal of Nonlinear Science , volume=. 2020 , publisher=

    work page 2020

  21. [21]

    Political studies , volume=

    Rethinking populism: Politics, mediatisation and political style , author=. Political studies , volume=. 2014 , publisher=

    work page 2014

  22. [22]

    Handbook of risk theory , pages=

    Cultural cognition as a conception of the cultural theory of risk , author=. Handbook of risk theory , pages=. 2012 , publisher=

    work page 2012

  23. [23]

    Proceedings of the International AAAI Conference on Web and Social Media , volume=

    Can-PolNews: A Multi-Platform Dataset of Political Discourse in Canada , author=. Proceedings of the International AAAI Conference on Web and Social Media , volume=

    work page

  24. [24]

    The Journal of Politics , volume=

    Political sophistication and the dimensionality of elite and mass attitudes, 1980- 2004 , author=. The Journal of Politics , volume=. 2015 , publisher=

    work page 1980

  25. [25]

    European Union Politics , volume=

    The Russian threat and the consolidation of the West: How populism and EU-skepticism shape party support for Ukraine , author=. European Union Politics , volume=. 2024 , publisher=

    work page 2024

  26. [26]

    Advances in Neural Information Processing Systems , volume=

    Mind the gap: Assessing temporal generalization in neural language models , author=. Advances in Neural Information Processing Systems , volume=

    work page

  27. [27]

    Physical review , volume=

    On the theory of the Brownian motion , author=. Physical review , volume=. 1930 , publisher=

    work page 1930

  28. [28]

    2012 , publisher=

    Machine learning: a probabilistic perspective , author=. 2012 , publisher=

    work page 2012

  29. [29]

    Party Politics , volume=

    Changing positions without losing face: How parties shift their position on issues , author=. Party Politics , volume=. 2023 , publisher=

    work page 2023

  30. [30]

    1998 , publisher=

    Private truths, public lies: The social consequences of preference falsification , author=. 1998 , publisher=

    work page 1998

  31. [31]

    Opinion dynamics: Statistical physics and beyond

    Opinion dynamics: Statistical physics and beyond , author=. arXiv preprint arXiv:2507.11521 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  32. [32]

    2023 , publisher=

    The canadian information ecosystem , author=. 2023 , publisher=

    work page 2023

  33. [33]

    Current biology , volume=

    Bistability, bifurcations, and Waddington's epigenetic landscape , author=. Current biology , volume=. 2012 , publisher=

    work page 2012

  34. [34]

    Journal of Experimental Social Psychology , volume=

    Attitude importance and attitude change , author=. Journal of Experimental Social Psychology , volume=. 1988 , publisher=

    work page 1988

  35. [35]

    Computer physics communications , volume=

    The calculation of the potential of mean force using computer simulations , author=. Computer physics communications , volume=. 1995 , publisher=

    work page 1995

  36. [36]

    The Journal of Politics , volume=

    Flip-flopping and electoral concerns , author=. The Journal of Politics , volume=. 2021 , publisher=

    work page 2021

  37. [37]

    Machine Learning: Science and Technology , year=

    Deeptime: a Python library for machine learning dynamical models from time series data , author=. Machine Learning: Science and Technology , year=

    work page

  38. [38]

    Political Psychology , volume=

    Understanding the determinants of political ideology: Implications of structural complexity , author=. Political Psychology , volume=. 2014 , publisher=

    work page 2014

  39. [39]

    Proceedings of the International AAAI Conference on Web and Social Media , volume=

    Online social behavior enhanced detection of political stances in tweets , author=. Proceedings of the International AAAI Conference on Web and Social Media , volume=

    work page

  40. [40]

    Proceedings of the International AAAI Conference on Web and Social Media , volume=

    Doubleh: Twitter user stance detection via bipartite graph neural networks , author=. Proceedings of the International AAAI Conference on Web and Social Media , volume=

    work page

  41. [41]

    Proceedings of the Association for Information Science and Technology , volume=

    Exploring the Opinion Shifting of Social Media Users Based on Stance Detection , author=. Proceedings of the Association for Information Science and Technology , volume=. 2024 , publisher=

    work page 2024

  42. [42]

    International Conference on Discovery Science , pages=

    From Tweets to Stance: An unsupervised framework for user stance detection on twitter , author=. International Conference on Discovery Science , pages=. 2023 , organization=

    work page 2023

  43. [43]

    PloS one , volume=

    Social context in political stance detection: Impact and extrapolation , author=. PloS one , volume=. 2025 , publisher=

    work page 2025

  44. [44]

    arXiv preprint arXiv:2412.02712 , year=

    Analyzing political stances on Twitter in the lead-up to the 2024 US election , author=. arXiv preprint arXiv:2412.02712 , year=

    work page arXiv 2024

  45. [45]

    , author=

    Do needs for security and certainty predict cultural and economic conservatism? A cross-national analysis. , author=. Journal of personality and social psychology , volume=. 2014 , publisher=

    work page 2014

  46. [46]

    Communications of the ACM , volume=

    Overcoming the J-shaped distribution of product reviews , author=. Communications of the ACM , volume=. 2009 , publisher=

    work page 2009

  47. [47]

    Public opinion quarterly , volume=

    What do we measure when we measure affective polarization? , author=. Public opinion quarterly , volume=. 2019 , publisher=

    work page 2019

  48. [48]

    Health Psychology Review , volume=

    Attractor landscapes: A unifying conceptual model for understanding behaviour change across scales of observation , author=. Health Psychology Review , volume=. 2023 , publisher=

    work page 2023

  49. [49]

    , author=

    Attitudes as catastrophes: From dimensions to categories with increasing involvement. , author=. 1994 , publisher=

    work page 1994

  50. [50]

    Multivariate Behavioral Research , volume=

    From metaphor to computation: Constructing the potential landscape for multivariate psychological formal models , author=. Multivariate Behavioral Research , volume=. 2023 , publisher=

    work page 2023

  51. [51]

    Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) , pages=

    Zero-Shot Stance Detection: A Dataset and Model using Generalized Topic Representations , author=. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) , pages=

    work page 2020

  52. [52]

    Journal of Elections, Public Opinion and Parties , volume=

    Social conformity or attitude persistence? The bandwagon effect and the spiral of silence in a polarized context , author=. Journal of Elections, Public Opinion and Parties , volume=. 2024 , publisher=

    work page 2024

  53. [53]

    The Journal of Machine Learning Research , volume=

    Spectral regularization algorithms for learning large incomplete matrices , author=. The Journal of Machine Learning Research , volume=. 2010 , publisher=

    work page 2010

  54. [54]

    Biometrika , volume=

    Inference and missing data , author=. Biometrika , volume=. 1976 , publisher=

    work page 1976

  55. [55]

    Survey nonresponse , pages=

    The causes of no-opinion responses to attitude measures in surveys: They are rarely what they appear to be , author=. Survey nonresponse , pages=. 2002 , publisher=

    work page 2002

  56. [56]

    American Politics Research , volume=

    Public ideology and political dynamics in the United States , author=. American Politics Research , volume=. 2012 , publisher=

    work page 2012

  57. [57]

    Proceedings of the Twelfth Language Resources and Evaluation Conference , pages=

    French tweet corpus for automatic stance detection , author=. Proceedings of the Twelfth Language Resources and Evaluation Conference , pages=

    work page

  58. [58]

    Proceedings of the Twelfth Language Resources and Evaluation Conference , pages=

    Multilingual stance detection in tweets: The Catalonia independence corpus , author=. Proceedings of the Twelfth Language Resources and Evaluation Conference , pages=

    work page

  59. [59]

    2023 IEEE International Conference on Data Mining (ICDM) , pages=

    Contextual target-specific stance detection on twitter: Dataset and method , author=. 2023 IEEE International Conference on Data Mining (ICDM) , pages=. 2023 , organization=

    work page 2023

  60. [60]

    Proceedings of the 32nd ACM International Conference on Multimedia , pages=

    Multimodal Multi-turn Conversation Stance Detection: A Challenge Dataset and Effective Model , author=. Proceedings of the 32nd ACM International Conference on Multimedia , pages=

    work page

  61. [61]

    Proceedings of the 10th international workshop on semantic evaluation (SemEval-2016) , pages=

    Semeval-2016 task 6: Detecting stance in tweets , author=. Proceedings of the 10th international workshop on semantic evaluation (SemEval-2016) , pages=

    work page 2016

  62. [62]

    Findings of the association for computational linguistics: ACL-IJCNLP 2021 , pages=

    P-stance: A large dataset for stance detection in political domain , author=. Findings of the association for computational linguistics: ACL-IJCNLP 2021 , pages=

    work page 2021

  63. [63]

    Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    EZ-STANCE: A large dataset for English zero-shot stance detection , author=. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    work page

  64. [64]

    2025 , eprint=

    Qwen3 Technical Report , author=. 2025 , eprint=

    work page 2025

  65. [65]

    arXiv preprint arXiv:2506.10942 , year=

    Building a Media Ecosystem Observatory from Scratch: Infrastructure, Methodology, and Insights , author=. arXiv preprint arXiv:2506.10942 , year=

    work page arXiv

  66. [66]

    Breakthroughs in statistics: Methodology and distribution , pages=

    Statistical methods for research workers , author=. Breakthroughs in statistics: Methodology and distribution , pages=. 1970 , publisher=

    work page 1970

  67. [67]

    kdd , volume=

    A density-based algorithm for discovering clusters in large spatial databases with noise , author=. kdd , volume=

    work page

  68. [68]

    Annalen der physik , volume=

    On the motion of small particles suspended in liquids at rest required by the molecular-kinetic theory of heat , author=. Annalen der physik , volume=

    work page

  69. [69]

    Physica D: nonlinear phenomena , volume=

    Determining Lyapunov exponents from a time series , author=. Physica D: nonlinear phenomena , volume=. 1985 , publisher=

    work page 1985

  70. [70]

    2018 , publisher=

    Forecasting: principles and practice , author=. 2018 , publisher=

    work page 2018

  71. [71]

    Econometrica: Journal of the Econometric Society , pages=

    Heteroskedasticity and autocorrelation consistent covariance matrix estimation , author=. Econometrica: Journal of the Econometric Society , pages=. 1991 , publisher=

    work page 1991

  72. [72]

    The Fokker-Planck equation: methods of solution and applications , pages=

    Fokker-planck equation , author=. The Fokker-Planck equation: methods of solution and applications , pages=. 1989 , publisher=

    work page 1989

  73. [73]

    Journal of the American statistical association , volume=

    Distribution of the estimators for autoregressive time series with a unit root , author=. Journal of the American statistical association , volume=. 1979 , publisher=

    work page 1979

  74. [74]

    arXiv preprint arXiv:2201.01322 , year=

    Opinion dynamics in social networks: From models to data , author=. arXiv preprint arXiv:2201.01322 , year=

    work page arXiv

  75. [75]

    Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    Predicting opinion dynamics via sociologically-informed neural networks , author=. Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    work page

  76. [76]

    Proceedings of the National Academy of Sciences , volume=

    The nonlinear feedback dynamics of asymmetric political polarization , author=. Proceedings of the National Academy of Sciences , volume=. 2021 , publisher=

    work page 2021

  77. [77]

    Journal of complex networks , volume=

    The polarization within and across individuals: The hierarchical Ising opinion model , author=. Journal of complex networks , volume=. 2020 , publisher=

    work page 2020

  78. [78]

    Annual review of psychology , volume=

    Attitude strength , author=. Annual review of psychology , volume=. 2017 , publisher=

    work page 2017

  79. [79]

    British Journal of Political Science , volume=

    Does counter-attitudinal information cause backlash? Results from three large survey experiments , author=. British Journal of Political Science , volume=. 2020 , publisher=

    work page 2020

  80. [80]

    Science advances , volume=

    The small effects of political advertising are small regardless of context, message, sender, or receiver: Evidence from 59 real-time randomized experiments , author=. Science advances , volume=. 2020 , publisher=

    work page 2020

Showing first 80 references.