PI: Ishanu Chattopadhyay, PhD

Assistant Professor of Biomedical Informatics & Computer Science

University of Kentucky

DARPA-EA-25-02-05-MAGICS-PA-025

HR0011-26-3-E016

May 2026

Summary of Scientific Accomplishments

For public dissemination

• A general modeling stack now exists for creating digital twins of opinion surveys, with detailed oos validation netrics

1. Large Science Models: Latent dependency non-parametric inference model working code (https://github.com/zeroknowledgediscovery/quasinet)

2. Extensive out-of-sample validation of response reconstruction showing substantial improvement over LLM based survey models (Fig 2)

Fig. 1

Fig. 2

LLM model: GPT 5.4

Baseline: independent generation

Fig. 3

3. Digital Twin Anchored Generation (DTAG) framework code in beta testing. Allowing a LLM semantic layer to access LSM models stratified by time, geolocation and demographics to make it possible to have human "conversations" with persona without hallucinations.

3. With DTAG, Targeted question sequences being investigated to drive opinion changes (Fig.3)

4. D. Yang, J. Evans and I. Chattopadhyay, “The Fabric of Opinion: How Interdependent Attitudes Constrain Worldviews, Predict Behavior, and Evolve Over Time", Nature Human Behavior, Under Review

Planned Abstract Submission:

Accepted Paper:

K. Wu, F. Li and I. Chattopadhyay, "Emergenet: A Digital Twin of Influenza A Evolution for Vaccine Strain Forecasting and Emergence Risk Assessment",  Military Medicine, In Press

DTAG: Global Digital Twin of Opinions

Recall Bail etal.

“Exposure to opposing views on social media can increase political polarization” by Christopher A. Bail et al., published in PNAS in September 2018 (Vol. 115, No. 37, pp. 9216–9221; DOI: 10.1073/pnas.1804840115)

We find more general possibilities: We can make world-views go more extreme or less extreme based on the line of questions and the persona

Perturbing with opposing views made conservatives more conservative (statistically significant), liberals more liberal (not statistically significant)

LSM

• Fixed Question Sets Exist That Move Different Persona Towards Polarization/Depolarization

• We can optimize question sequences to move the same persona in a chosen direction

• Note: It is relatively easy to polarize than to depolarize

Prospective validation in Human Cohorts

• Beyond MAGICS Scope
• Prolific Experiments using non-DARPA External Funding with UKy IRB approval

x

demographic filter

persona filter

P2

P1

fixed question sets

attention questions

Milestones                           

\(\checkmark\)

\(\checkmark\)

LARGE SCIENCE MODELS · SYSTEM-LEVEL DRIFT

The Characteristic Drift Spectrum

A task-free, multiscale fingerprint of systemic organization in complex systems

Structural drift

Information drift

Kolmogorov drift

Ishanu Chattopadhyay

University of Kentucky

MOTIVATION

Why a characteristic drift spectrum?

Characteristic Drift Spectra of Complex Systems

2 / 14

Classical dataset & concept drift

Task model / classifier

• Tied to a specified prediction problem, response variable, or global distributional comparison
• Classifier two-sample tests use discriminator performance as evidence of change
• Inherits an arbitrary choice of response or classifier

Characteristic drift spectrum

depth 0

depth 1

depth 2

• Response-variable-free: no privileged Y is required
• Maps high-dimensional conditional organization to a few scalar probes
• Retains a structural-scale coordinate, so change is located in depth

The question becomes: where in the hierarchy of system states does organization change, and in what way?

LARGE SCIENCE MODEL

An LSM encodes conditional organization

Characteristic Drift Spectra of Complex Systems

3 / 14

For categorical state X = (X1, …, Xp) and row population R, an LSM is a target-wise collection of conditional predictors:

For row x, the mean conditional log-persistence is

PM(x) is the average log support assigned to the observed coordinates when each coordinate is predicted from the remainder of the state.

Low persistence identifies rows that are poorly reconstructed by the dominant learned conditional organization.

LSM construction: Sizemore et al., Sci. Adv. 2024; Wu, Li & Chattopadhyay, Mil. Med. 2026.

Masked-coordinate prediction

x1

x2

x3

?

masked

⋯

xp

MAP

observed

Conditional distribution over states a ∈ Ai

INTRINSIC RESIDUAL HIERARCHY

Peeling recovers a residual hierarchy

Characteristic Drift Spectra of Complex Systems

4 / 14

⋮

a fresh LSM at every level

1

Fit and score

2

Robustly standardize lack of persistence

3

Peel and refit

At depth d, fit Md = LSM(Rd) and compute the persistence of every row under Md.

Intuition. Peeling removes the statistically dominant core and refits on the surviving low-persistence population, so organization masked at the population level becomes visible. Depth is structural, not chronological.

COORDINATES OF THE SPECTRUM

Three scalar probes at every depth

Characteristic Drift Spectra of Complex Systems

5 / 14

PROBE I

Structural drift

MEASURES

Change in MAP-normalized masked-state reconstruction support after refitting.

BUILT FROM

The MAP-normalized reconstruction score of the LSM

PROBE II

Information drift

MEASURES

Change in description length under the conditional structure learned from the data.

BUILT FROM

The data-adaptive code length induced by the LSM

PROBE III

Kolmogorov drift

MEASURES

Compression-calibrated information displacement between adjacent residual levels.

BUILT FROM

Normalized compression distance (gzip) between matched row blocks

At every depth d, the three numbers form one point of the spectrum:

PROBE I

Structural drift

Characteristic Drift Spectra of Complex Systems

6 / 14

MAP-normalized masked-state reconstruction score

uM = 1 when the observed state attains the conditional MAP support; smaller values measure its support relative to that MAP state.

With a balanced probe cloud Qd and common evaluable targets Id(x), average the change between adjacent depths:

Same probe, adjacent depths

MAP state

observed state

Intuition. Structural drift measures how the system’s learned ability to reconstruct masked coordinates changes across structural depth.

PROBE II

Information drift

Characteristic Drift Spectra of Complex Systems

7 / 14

A data-adaptive code from the learned conditional structure

Mean code length (bits) of row x when each coordinate is encoded with the LSM’s conditional predictor.

Information drift between adjacent residual levels

Both expectations use the depth-d model Md: rows of Rd+1 are coded with structure learned on Rd.

Mean code length under Md

schematic

Intuition. An LSM-based description-length probe: information displacement is measured using structure learned from the system itself.

PROBE III

Kolmogorov drift

Characteristic Drift Spectra of Complex Systems

8 / 14

At depth d, draw matched row blocks A1, A2 ⊂ Rd and B1, B2 ⊂ Rd+1. For serialized blocks a, b and compressor length C(·), the normalized compression distance is

Within-depth distances establish the intrinsic compression scale; the across-depth distance measures separation between adjacent levels:

within depth: intrinsic scale

across depth: separation

Intuition. Compression-calibrated information displacement between adjacent residual levels, measured with a general-purpose compressor rather than the learned model.

SYNTHESIS

The characteristic drift spectrum

Characteristic Drift Spectra of Complex Systems

9 / 14

At every depth d, the three numbers form

and stacking them across depth gives a depth-by-coordinate matrix:

⋮

⋮

⋮

rows: structural depth d = 0, …, D−1 columns: the three probes

A high-dimensional system is mapped to a small, ordered set of scalar probes while preserving structural depth: a task-free fingerprint that can be compared across repeated observations.

POSITIVE CONTROL

GSS: recovering a known data shift

Characteristic Drift Spectra of Complex Systems

10 / 14

0.0750

2012–2018 mean · face-to-face

waves: 0.0738, 0.0749, 0.0764, 0.0749

0.1531

2021 · primarily web-based

≈ 2× the pre-2021 level

0.1420 → 0.1117

2022, 2024 · mixed web/face-to-face

partial return toward the earlier regime

The spectrum recovers this independently documented change in the observation process without being trained to discriminate survey years or modes.

POSITIVE CONTROL

Two projections of the GSS spectrum

Characteristic Drift Spectra of Complex Systems

11 / 14

a Polar rendering of the raw (information drift, structural drift) plane. A tight 2012–2018 regime, a pronounced 2021 displacement, and partial return in 2022 and 2024.

b gzip/NCD information displacement against the Dirichlet-posterior mean of structural drift; the horizontal axis shows exp(Kolmogorov drift).

MULTISCALE COMPARISON

Spectra as fingerprints: 2018 vs 2024

Characteristic Drift Spectra of Complex Systems

12 / 14

d = 0

.0971

.0749

.00061

d = 1

.1130

.0460

.00004

d = 2

.1164

.0395

−.00090

d = 0

.1165

.1117

.00015

d = 1

.1383

.0754

−.00008

d = 2

.1472

.0794

−.00035

First three transitions; cell shading scales with value within each of the first two columns.

Reading the fingerprints

• Shallowest transition: information drift rises from .0749 (2018) to .1117 (2024).
• 2024 exceeds 2018 in both structural and information drift at all three depths.
• Structural drift grows with depth in both years; Kolmogorov drift stays below 0.001 in magnitude.
• Repeated observations can be compared through their full spectra rather than a single classifier or one global distance.

OUTLOOK

Interpretation and next steps

Characteristic Drift Spectra of Complex Systems

13 / 14

A system-level measurement object: a high-dimensional complex system mapped to a small, ordered set of scalar probes, with structural depth preserved.

Applies whenever a multivariate system admits conditional modeling without selecting a privileged response variable.

ROUTES FOR CALIBRATING THE OPERATING RANGE OF EACH COORDINATE

Matched-size refit nulls

Refit at matched sample sizes to separate real drift from refit variability.

Independent peeling rules

Recover the hierarchy with alternative peeling rules to test robustness.

Planted-change simulations

Inject known changes to map the sensitivity of each coordinate.

Alternative compressors

Test how Kolmogorov drift depends on the choice of gzip.

Observation-design controls

Separate changes in how data are collected from changes in the system.

ACKNOWLEDGMENT & REFERENCES

References

Characteristic Drift Spectra of Complex Systems

14 / 14

[1] J. G. Moreno-Torres et al., “A unifying view on dataset shift in classification,” Pattern Recognition 45(1):521–530, 2012.

[2] J. Gama et al., “A survey on concept drift adaptation,” ACM Computing Surveys 46(4), 2014.

[3] A. Gretton et al., “A kernel two-sample test,” JMLR 13:723–773, 2012.

[4] D. Lopez-Paz and M. Oquab, “Revisiting classifier two-sample tests,” ICLR, 2017.

[5] N. Sizemore et al., “A digital twin of the infant microbiome to predict neurodevelopmental deficits,” Science Advances 10(15):eadj0400, 2024.

[6] K. Wu, F. Li, and I. Chattopadhyay, “Emergenet: A digital twin of influenza A evolution for vaccine strain forecasting and emergence risk assessment,” Military Medicine, usag255, 2026.

[7] NORC at the University of Chicago, “General Social Survey 2021 codebook,” 2021.

[8] NORC, “General Social Survey 2022 codebook,” 2023.

[9] NORC, “General Social Survey 2024 codebook,” release 2, 2025.

[10] NORC, “Sequential data collection designs in the General Social Survey,” GSS Methodological Report 140, 2025.

[11] M. Li et al., “The similarity metric,” IEEE Trans. Information Theory 50(12):3250–3264, 2004.

Correspondence: ishanu_ch@uky.edu

Acknowledgment. This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) under Agreement No. HR0011-26-3-E016. The views, opinions, and/or findings expressed are those of the authors and should not be interpreted as representing the official views or policies of the Department of Defense or the U.S. Government.