MétaCan
Menu
Back to cohort

Bridging the Generation Gap: Age-Parity in Synthetic Customer Profiles

2025· article· W4417249539 on OpenAlexaff
Nitin Kumar, Naga Satya Praveen Kumar Yadati, Shiva Kumar Ramavathr

Bibliographic record

Venuenot available
Typearticle
Language
FieldComputer Science
TopicRecommender Systems and Techniques
Canadian institutionsMarriott International (Canada)
Fundersnot available
KeywordsBridging (networking)AutoencoderBaseline (sea)Transaction dataPersonalizationSynthetic dataDiscriminative modelAnalyticsBenchmark (surveying)

Abstract

fetched live from OpenAlex

Synthetic customer profiles are increasingly used to augment limited datasets for marketing analytics and personalization systems. Yet, generative methods often amplify imbalances between demographic segments-particularly younger <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\leq 35)$</tex> and older (>35) customers-leading to biased downstream models. We propose a Fairness-Constrained Tabular Variational Autoencoder (FC-TVAE) that enforces age-band demographic parity during profile generation. Starting with the publicly available Kaggle Customer Personality Analysis dataset (2,240 records with age, marital status, income, and annual spend across six product categories), we min-max normalize continuous features and discretize age into two bands <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\leq 35$</tex> vs. <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$&gt;35)$</tex>. A maximum-mean-discrepancy regularizer is applied to the VAE's latent distributions to minimize the normalized demographic-parity gap (DPG) between age slices. We benchmark FC-TVAE against Baseline TVAE, Conditional GAN, FairGAN, and DP-WGAN, using five random seeds. On the normalized spend scale, FC-TVAE achieves an MSE of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.0158 \pm 0.0004(\approx 12 \%$</tex> average absolute error), compared to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.0150-0.0170$</tex> for baselines. Crucially, FCTVAE reduces DPG to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.055 \pm 0.006$</tex>, a 27% improvement over FairGAN <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(0.075 \pm 0.006)$</tex> and 42% over Conditional GAN <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(0.095 \pm \mathbf{0. 0 0 7}) (p&lt;\mathbf{0. 0 1})$</tex>. Minority-group recall for the younger cohort rises to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{0. 7 5} \pm \mathbf{0. 0 2}$</tex>, outperforming FairGAN <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathbf{0. 6 8} \pm \mathbf{0. 0 2})$</tex> by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 0}$</tex>% and DP-WGAN <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(0.66 \pm 0.03)$</tex> by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$13.6 \%(p&lt;0.005)$</tex>. These results demonstrate that integrating age-band parity constraints at the model level yields the best balance of fidelity, fairness, and coverage. FC-TVAE offers a practical, extensible framework for generating inclusive synthetic profiles, enabling more equitable personalization across age groups in marketing analytics.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.002
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: none
Teacher disagreement score0.979
Threshold uncertainty score0.917

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0020.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.001
Science and technology studies0.0000.000
Scholarly communication0.0010.000
Open science0.0010.001
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.043
GPT teacher head0.291
Teacher spread0.248 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one teacher head, not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations0
Published2025
Admission routes1
Has abstractyes

Explore more

Same topicRecommender Systems and TechniquesFrench-language works237,207