pith. sign in

arxiv: 2509.11982 · v2 · submitted 2025-09-15 · 💻 cs.LG · cs.CY

Examining the Relationship between Scientific Publishing Activity and Hype-Driven Financial Bubbles: A Comparison of the Dot-Com and AI Eras

Aksheytha Chelikavada , Casey C. Bennett This is my paper

Pith reviewed 2026-05-18 15:50 UTC · model grok-4.3

classification 💻 cs.LG cs.CY
keywords financial bubblesscientific publishingcitation networksdot-com eraAI erasocial network analysishype signalsmarket prediction
0
0 comments X

The pith

Citation network patterns from the dot-com era do not predict an AI financial bubble.

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

This paper examines whether signals in scientific publication citation networks can forecast financial bubbles by comparing the dot-com era, which ended in a market crash, to the current AI era. Researchers tracked temporal changes in how scientists influence one another through citations during both periods using social network analysis. They found that overall patterns tied to the dot-com bubble do not clearly emerge in AI research, even though some AI scientists display similar influence dynamics. Predictive models including LSTM, KNN, and ARX/GARCH applied to the data indicate either no bubble in AI or one of a form not seen before. The results suggest that historical publishing activity may not reliably forecast outcomes in today's technology markets.

Core claim

The paper establishes that temporal social network analysis of publication citation networks during the dot-com era from 1994 to 2001 does not provide definitive predictions for the rise and fall of a financial bubble in the AI era from 2017 to 2024. Although there are observable changes in yearly citation networks reflecting shifts in scientists' publishing behavior, and a subset of AI-era scientists exhibit publication influence patterns similar to those in the dot-com period, analyses using LSTM, KNN, and ARX/GARCH models point toward either an unprecedented type of financial bubble or the absence of a bubble altogether in the AI market. The conclusion is that patterns from the dot-com er

What carries the argument

Temporal social network analysis (SNA) on publication citation networks, which tracks evolving influence and connectivity among scientists over time to detect potential hype signals connected to market movements.

If this is right

  • Dot-com era citation patterns cannot be used to forecast the trajectory of AI-related financial markets.
  • A subset of AI scientists shows influence dynamics similar to dot-com researchers, indicating some continuity in research hype.
  • Machine learning models like LSTM and KNN applied to network data fail to detect clear bubble signals in the AI era.
  • The AI era may involve different mechanisms driving potential market disruptions compared to past technology bubbles.
  • Scientific publishing activity alone may not capture the full set of factors leading to financial bubbles in emerging tech.

Where Pith is reading between the lines

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

  • If the AI era lacks a traditional bubble, sustained scientific progress in AI could support more stable market growth than occurred in the dot-com period.
  • Extending this citation network analysis to other fields like quantum computing or biotechnology could test whether publishing patterns predict bubbles more broadly.
  • Future work might combine citation data with venture capital funding trends to improve detection of market hype.
  • The mirroring patterns in a subset of scientists suggest that certain research communities may carry over hype dynamics across different technology eras.

Load-bearing premise

The dot-com and AI eras are similar enough that citation networks in scientific publishing reflect comparable hype levels driving financial bubbles, allowing patterns from one to be tested in the other.

What would settle it

If AI-related stock markets undergo a sharp crash matching the dot-com pattern while citation networks display the same temporal shifts and mirroring influence dynamics seen in the 1990s, this would contradict the finding that those patterns do not predict AI bubbles.

Figures

Figures reproduced from arXiv: 2509.11982 by Aksheytha Chelikavada, Casey C. Bennett.

Figure 6
Figure 6. Figure 6: Investigating the results in [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
read the original abstract

Financial bubbles often arrive without much warning, but create long-lasting economic effects. For example, during the dot-com bubble, innovative technologies created market disruptions through excitement for a promised bright future. Such technologies originated from research where scientists had developed them for years prior to their entry into the markets. That raises a question on the possibility of analyzing scientific publishing data (e.g. citation networks) leading up to a bubble for signals that may forecast the rise and fall of similar future bubbles. To that end, we utilized temporal SNAs to detect possible relationships between the publication citation networks of scientists and financial market data during two modern eras of rapidly shifting technology: 1) dot-com era from 1994 to 2001 and 2) AI era from 2017 to 2024. Results showed that the patterns from the dot-com era (which did end in a bubble) did not definitively predict the rise and fall of an AI bubble. While yearly citation networks reflected possible changes in publishing behavior of scientists between the two eras, there was a subset of AI era scientists whose publication influence patterns mirrored those during the dot-com era. Upon further analysis using multiple analysis techniques (LSTM, KNN, AR X/GARCH), the data seems to suggest two possibilities for the AI era: unprecedented form of financial bubble unseen or that no bubble exists. In conclusion, our findings imply that the patterns present in the dot-com era do not effectively translate in such a manner to apply them to the AI market.

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

2 major / 2 minor

Summary. The manuscript claims to examine the relationship between scientific publishing activity via citation networks and hype-driven financial bubbles by comparing the dot-com era (1994-2001) and AI era (2017-2024) using temporal social network analysis (SNA) and models such as LSTM, KNN, and ARX/GARCH. It concludes that patterns from the dot-com era, which ended in a bubble, do not definitively predict the rise and fall of an AI bubble, suggesting either an unprecedented bubble or no bubble, while identifying a subset of AI scientists with mirroring publication influence patterns.

Significance. If the results hold after addressing methodological gaps, the paper would provide evidence that historical bubble patterns may not generalize to the AI era due to evolving publishing behaviors. This could have implications for using scientometric data in financial forecasting and highlight the need for era-specific models in bubble detection.

major comments (2)
  1. [Abstract and Methods] The abstract and methods sections provide no quantitative details on data sources for citation networks and financial data, model specifications, hyperparameters for LSTM, KNN, and ARX/GARCH, validation steps, or error measures. This makes the support for the no-prediction claim unverifiable.
  2. [Discussion of SNA patterns] The analysis does not control for scale and structural differences in publication volumes, network density, citation practices, or incentive structures between the dot-com and AI eras when comparing SNA metrics as hype proxies. Without normalization, differences in temporal patterns or model outputs could stem from data artifacts rather than bubble mechanisms, which is load-bearing for the central claim that dot-com patterns do not predict AI bubble dynamics.
minor comments (2)
  1. [Abstract] There is a typographical error: 'AR X/GARCH' should likely be 'ARX/GARCH'.
  2. [Abstract] The abstract mentions 'multiple analysis techniques' but does not reference specific results, tables, or figures to support the findings.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight important areas for improving methodological transparency and robustness in our comparison of citation network patterns across the dot-com and AI eras. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract and Methods] The abstract and methods sections provide no quantitative details on data sources for citation networks and financial data, model specifications, hyperparameters for LSTM, KNN, and ARX/GARCH, validation steps, or error measures. This makes the support for the no-prediction claim unverifiable.

    Authors: We agree that additional quantitative details are needed for verifiability. The abstract was kept concise, and the methods section in the current version focuses on high-level description of temporal SNA and the listed models without full specifications. In the revised manuscript we will expand both sections to report the precise data sources (citation database and financial indices), model architectures, hyperparameter values, validation approach (e.g., rolling-window time-series cross-validation), and error metrics (RMSE, MAE, and classification accuracy where applicable). These additions will directly support evaluation of the no-prediction result. revision: yes

  2. Referee: [Discussion of SNA patterns] The analysis does not control for scale and structural differences in publication volumes, network density, citation practices, or incentive structures between the dot-com and AI eras when comparing SNA metrics as hype proxies. Without normalization, differences in temporal patterns or model outputs could stem from data artifacts rather than bubble mechanisms, which is load-bearing for the central claim that dot-com patterns do not predict AI bubble dynamics.

    Authors: We acknowledge that direct comparison of raw SNA metrics across eras with substantially different publication volumes and network densities risks confounding. Our models (LSTM, KNN, ARX/GARCH) were applied to normalized temporal sequences within each era, and the central claim rests on the failure of dot-com-trained models to generalize to the AI period rather than on absolute metric values. Nevertheless, to isolate bubble-related signals more cleanly, the revision will add explicit normalization steps (e.g., per-node or per-edge scaling of centrality and density measures) together with a sensitivity analysis showing that the lack of predictive transfer persists after these adjustments. We will also briefly discuss known shifts in citation practices and incentives between the two periods. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical comparison of external datasets

full rationale

The paper extracts temporal SNA metrics from citation networks and financial time series for two distinct eras, then applies LSTM/KNN/ARX-GARCH models to test whether dot-com patterns forecast AI-era behavior. All inputs are external publication and market data; no parameter is fitted to a subset and then relabeled as a prediction, no self-citation supplies a uniqueness theorem, and no ansatz or renaming is smuggled in. The conclusion that dot-com patterns do not definitively predict an AI bubble follows directly from the observed differences in network evolution and model outputs without reducing to the inputs by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim depends on the domain assumption that citation network dynamics serve as a proxy for hype relevant to bubbles, plus implicit choices in the machine learning models; no new entities are postulated.

free parameters (1)
  • LSTM, KNN, and ARX/GARCH model parameters
    The abstract invokes these models for analysis, which typically require hyperparameter selection or fitting to the specific datasets.
axioms (1)
  • domain assumption Temporal changes in publication citation networks reflect shifts in scientific influence and hype that can relate to financial market bubbles
    This premise underpins the use of temporal SNAs to compare eras and test for predictive signals.

pith-pipeline@v0.9.0 · 5814 in / 1462 out tokens · 75270 ms · 2026-05-18T15:50:19.784561+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

5 extracted references · 5 canonical work pages · 1 internal anchor

  1. [1]

    Bornmann, L., & Mutz, R. (2015). Growth rates of modern science: A bibliometric analysis based on the number of publications and cited references. Journal of the Association for Information Science and Technology, 66(11), 2215-2222. [11] Burt, R. S. (2001). Attachment, decay, and social network. Journal of Organizational Behavior: The International Journa...

    work page doi:10.1007/s13347-024-00817-w 2015

  2. [2]

    Maddodi, S., & Kunte, S. R. (2024). Market resilience in turbulent times: a proactive approach to predicting stock market responses during geopolitical tensions. Journal of Capital Markets Studies, 8(2), 173-194. [33] Malcom, D. (2018). It’s time we fix the peer review system. American Journal of Pharmaceutical Education, 82(5). [34] Marin, A., & Wellman,...

    work page doi:10.2139/ssrn.1152412 2024

  3. [3]

    Park, M., Leahey, E., & Funk, R. J. (2023). Papers and patents are becoming less disruptive over time. Nature, 613(7942), 138-144. [45] Partha, D., & David, P. A. (1994). Toward a new economics of science. Research policy, 23(5), 487-521. [46] Pasquinelli, M. (2023). The eye of the master: A social history of artificial intelligence. Verso Books. [47] Pix...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.2139/ssrn.1596024 2023

  4. [4]

    (2024, May)

    Steck, H., Ekanadham, C., & Kallus, N. (2024, May). Is cosine-similarity of embeddings really about similarity?. In Companion Proceedings of the ACM Web Conference 2024 (pp. 887-890). [58] Stephan, P. E. (1996). The economics of science. Journal of Economic Literature, 34(3), 1199-1235. [59] Sucarrat, G. (2019). The log-GARCH model via ARMA representation...

    work page doi:10.1016/j.ifacol.2015.11.031 2024

  5. [5]

    Waltman, L., & van Eck, N. J. (2012). A new methodology for constructing a publication-level classification system of science. Journal of the American Society for Information Science and Technology, 63(12), 2378–2392. https://doi.org/10.1002/asi.22748 [70] Watson, M. (1999). Rethinking capital mobility, re‐regulating financial markets. New Political Econo...

    work page doi:10.1002/asi.22748 2012