Health and AI Lab

Translating data into actionable knowledge

About us

We aim to improve human health, through the development of artificial intelligence methods. Most of these problems come back to the question of why things happen or how they change, so we focus on causal inference and time series data. We look at both clinical data as well as data generated outside of hospitals and aim to support both medical providers and patients in their decision making. Key application areas include stroke and diabetes. We are also working on devices that can automatically measure food intake, using body-worn sensors.


AI for precision nutrition center

We are thrilled to be part of the NIH's new nutrition for precision health program and leading the causal analysis of this incredible new data with $1.3 million in funding. Read more.

NIH R01 second renewal

Our R01 was renewed for another four years and $1.1 million! We're developing data driven simulations and causal inference methods to gain insight into consciousness and neurological status in neonatal and neuro ICUs.

Tell Me Something I Don't Know

Re-assessing how much we know can help us better use causal information when making decisions. Full paper published at CogSci 2020.

Better BG forecasting

In recent papers we show how patient generated data (from OpenAPS) can be used to develop robust forecasting models. We compare multiple methods and data types in an MLHC paper, and apply these to the BGLP forecasting challenge here.

How Causal Information Affects Decisions

Our paper published in CRPI finds that what we already know (and think we know) can prevent us from successfully using new information to make everyday decisions. NIH Research Matters covered this work.

AMIA 2019 Best Paper

We received the Homer R. Warner Award for our work on identifying new indicators for consciousness in NICU patients.

$2.3 million in new funding for our work

We received an NIH R01, NSF Smart & Connected Health grant, and NSF III grant that will support developing generalizable methods to harness the power of patient generated data, improving shared decision-making by combining computing and cognitive science, and transforming how we evaluate and communicate the output of AI/ML.

Better meal detection from CGM data

We leverage simulation to find meals from CGM data, which could improve artificial pancreas algorithms. Summary of our work, and full article in JAMIA.

Sabbatical at UCL

Samantha is on sabbatical in the psychology department at University College London.

Stevens covers our dietary monitoring work

New article on our research working toward fully automated dietary monitoring.


We are grateful for the support of multiple sponsors, including NSF, NIH, and the James S. McDonnell Foundation.