The first step is getting your precious, precious data!
You automatically generate a lot of data exhaust, like receipts for supplements food prescriptions health records, labs, health apps, and wearables.
Unfortunately, it’s kind of worthless when it’s scattered all over the place and just being used by advertisers to target you with Viagra ads.
So we need a Digital Twin Safe where it can be safely imported and analyzed.
The FDAi Should Make Data Collection Effortless by Any Means
Import from Wearables and Apps
We’re working to make it effortless to import your data from lots of apps and wearable devices like physical activity, sleep, environmental factors, and vital signs.
Browser AI Agents
Autonomous AI agents could use your browser to collect your diet data from services like Instacart, supplement purchases from Amazon, CVS prescriptions, Quest lab results, etc.
Manual Data Collection
You should be able to easily record any symptom severities, foods, treatments, or anything in a simple reminder inbox.
Data from Speech
If you prefer talking over messing around with apps, we’re working to make it possible to talk to the FDAi. This would also allow passive inference of cognitive and emotional data.
Images to Data
You should be able to easily record any symptom severities, foods, treatments, or anything in a simple reminder inbox.
Notifications
Notifications with action buttons can be used to track symptoms and factors in a fraction of a second.
Data
Analysis
After the FDAi gets a few months of your data, it should then analyze it and generate N-of-1 personal studies telling you how much different medications, supplements, or foods might improve or worsen your symptoms:
Causal
Inference
As any obnoxious college graduate will tell you, correlation does not necessarily imply causation.
Just because you took a drug and got better it doesn’t mean that’s really why your symptoms went away. Even in randomized controlled trials hundreds of other things are changing in your life and diet.
So your FDAi has to to apply Hill’s 6 Criteria for Causality to try to infer if something causes a symptom to worsen or improve instead of just seeing what correlates with the change.
Personal
Studies
For instance, when gluten-sensitive people eat delicious gluten, it usually takes about a 2-day onset delay before they start having symptoms. Then, when they stop eating it, there’s usually a 10-day duration of action before their gut heals and their symptoms improve.
Onset Delays
One example of causal inference involves applying forward and reverse lagging of the depression and exercise data. The result suggests a causal relationship based on the temporal precedence of physical activity.
Discovering
Cumulative Effects
The FDAi should also compare the outcome over various durations following the exposure to see if there is a long-term cumulative effect or if it’s just a short-term acute effect. Acute effects are probably obvious already. This analysis suggests that the mood benefits of regular exercise may continue to accumulate after at least a month of above-average exercise.
