Back to AI Research

AI Research

Is It You or Your Environment? A Bayesian Inference... | AI Research

Key Takeaways

  • A Bayesian Inference Framework for Genomically-Anchored Personalized Physiological Interpretation Personalized health AI syste...
  • We propose a solution grounded in causal inference and Bayesian prior design.
  • The anchor initializes a Bayesian belief state over an individual's physiological set point G-hat = mu + sum(beta_i
  • g_i), where beta_i are GWAS-derived effect sizes and g_i are risk-allele counts.
  • Each incoming physiological measurement P produces a non-constitutional deviation delta = P - G-hat that separates the signal attributable to environment and state from the constitutionally fixed baseline.
Paper AbstractExpand

Personalized health AI systems face a fundamental cold-start problem: machine learning models for physiological interpretation require weeks of individual behavioral data before they can distinguish constitutional variation from environmentally driven deviation. We propose a solution grounded in causal inference and Bayesian prior design. An individual's genomic profile serves as an exogenous genetic anchor -- a domain-informed, personalized prior that is fixed at conception, immune to reverse causation, and available before a single behavioral observation is collected. The anchor initializes a Bayesian belief state over an individual's physiological set point G-hat = mu + sum(beta_i * g_i), where beta_i are GWAS-derived effect sizes and g_i are risk-allele counts. Each incoming physiological measurement P produces a non-constitutional deviation delta = P - G-hat that separates the signal attributable to environment and state from the constitutionally fixed baseline. As behavioral data accrue, the prior decays according to G-hat_t = w(t)*G-hat_genomic + [1-w(t)]*P-bar_t, transitioning from genome-dominated to empirical-baseline-dominated inference. The same observed HRV of 55 ms generates a suppression hypothesis for a person whose prior predicts 80 ms, and an enhancement hypothesis for a person whose prior predicts 30 ms -- a reversal impossible without a personalized anchor. We develop this architecture across six physiological domains, grading genomic priors by evidence strength, distinguishing robustly replicated anchors (FTO, FADS1/2, FKBP5) from contested candidate genes (SLC6A4, MAOA, DRD2). We address the inference boundary between association, Mendelian randomization, and individual token causation, and define four constraints for deployment: evidence-graded priors, dynamic decay, ancestry-matched effect sizes, and attribution rather than deterministic output.

Is It You or Your Environment? A Bayesian Inference Framework for Genomically-Anchored Personalized Physiological Interpretation
Personalized health AI systems often struggle with a "cold-start" problem: they require weeks of behavioral data to understand a user's unique physiology. Without this history, AI cannot distinguish between a person’s natural, constitutional baseline and a temporary change caused by their environment. This paper proposes a solution by using an individual's genomic profile as a "genetic anchor"—a fixed, personalized starting point available from day one. By combining this genetic information with Bayesian inference, the system can immediately begin interpreting physiological signals, such as heart rate or sleep patterns, relative to the individual’s own biology rather than generic population averages.

Bridging the Cold-Start Gap

Current health AI relies on either population norms or long-term personal behavioral baselines. Population norms are often misleading because they ignore individual differences, while personal baselines take too long to establish. The authors introduce an "exogenous genetic anchor," which is a set point derived from an individual's DNA. Because this genetic data is fixed at conception and cannot be changed by behavior or environment, it serves as a reliable, unbiased reference point. This allows the AI to initialize a "belief state" about a user's physiology before a single day of behavioral data is collected.

How the Causal Decomposition Works

The core of the framework is a causal decomposition layer that separates a person's physiological measurement into two parts: the constitutional baseline (the genetic anchor) and the non-constitutional deviation (the signal caused by environment, state, or behavior). For example, if two people have the same heart rate variability (HRV) reading of 55 ms, the AI uses their genetic anchors to interpret the result differently. If one person’s genetic expectation is 80 ms, the AI identifies a suppression of their autonomic tone. If another person’s expectation is 30 ms, the AI identifies an enhancement. This "normal for whom" approach allows for personalized causal hypotheses that would be impossible with standard population-based models.

Dynamic Updating and Uncertainty

The framework is designed to be flexible. As the system collects more behavioral data over time, it uses a "dynamic decay" process. Initially, the AI relies heavily on the genetic anchor, but as the user’s empirical behavioral baseline stabilizes, the system shifts its weight toward that real-world data. Furthermore, the authors emphasize "calibrated restraint." They grade genetic evidence by strength, acknowledging that some genetic markers (like FTO for metabolism) are robust, while others (like those for dopamine or serotonin) are contested or weak. The system uses an uncertainty model to ensure that low-evidence genetic priors do not lead to incorrect or spurious conclusions.

Important Considerations for Deployment

The authors stress that this framework is intended for attribution rather than deterministic output. They define four key constraints for any system using this approach:

  • Evidence-graded priors: Using only well-replicated genetic markers.

  • Dynamic decay: Gradually transitioning from genetic-based to behavior-based inference.

  • Ancestry-matched effect sizes: Ensuring that genetic data is interpreted correctly based on the individual's background.

  • Attribution-focused: Providing ranked causal hypotheses rather than claiming to provide definitive medical interventions.
    By following these constraints, the framework aims to provide a more accurate, personalized, and honest way to interpret health data from the very first day of monitoring.

Comments (0)

No comments yet

Be the first to share your thoughts!