Knowledge Vault 2/96 - ICLR 2014-2023
Elaine Nsoesie ICLR 2023 - Invited Talk - AI, History and Equity
<Resume Image >

Concept Graph & Resume using Claude 3 Opus | Chat GPT4 | Gemini Adv | Llama 3:

graph LR classDef equity fill:#f9d4d4, font-weight:bold, font-size:14px; classDef racialdata fill:#d4f9d4, font-weight:bold, font-size:14px; classDef health fill:#d4d4f9, font-weight:bold, font-size:14px; classDef neighborhood fill:#f9f9d4, font-weight:bold, font-size:14px; classDef policy fill:#f9d4f9, font-weight:bold, font-size:14px; A[Elaine Nsoesie
ICLR 2023] --> B[Equity: tailored solutions,
not just equality. 1] A --> C[AIM AHEAD, Racial Data
Tracker projects. 2] C --> D[Racial Data Tracker:
data on disparities. 3] A --> E[Neighborhood: major determinant
of life expectancy. 4] E --> F[Wealth, race intersect
with neighborhood. 4] A --> G[Social determinants impact
health more than care. 5] A --> H[US life expectancy varies
by race, gender. 6] A --> I[Redlining policies still
impact health, segregation. 7] A --> J[Machine learning estimates
obesity from imagery. 8] A --> K[Analysis: racial disparities in
built environment, health. 9] K --> L[164M images, 59K
neighborhoods analyzed. 10] K --> M[Neighborhoods classified by
race, redlining maps. 10] K --> N[Focus on health-related
built environment features. 11] K --> O[1930s policies still impact
racial composition. 12] K --> P[Housing quality mediates
sleep, asthma disparities. 13] A --> Q[Goal: measure inequities
to inform policy. 14] A --> R[Wealth, discrimination independently
shape neighborhoods. 15] A --> S[Analyzing future policy
impacts being considered. 16] A --> T[Change over time varies
by area, interventions. 17] A --> U[Data bias a concern,
must be mitigated. 18] A --> V[Targeted, community-engaged interventions
needed, not one-size-fits-all. 19] A --> W[Research impacts behavior,
outcomes consider in design. 20] A --> X[Quantifying disparities motivates
policymakers to change. 21] A --> Y[Collaborated with practitioners
to influence surveillance. 22] A --> Z[Research should provide
accessible policy guidance. 23] Z --> AA[Example: adding green space
to built environment. 24] A --> AB[Study other countries specific
policy contexts, impacts. 25] A --> AC[Examine policies with positive,
negative disparity outcomes. 26] A --> AD[1930s classifications problematic
avoid such labels. 27] A --> AE[Less segregation, better
socioeconomic health overall. 28] A --> AF[Racial disparities in
housing persist today. 29] A --> AG[BMI limitations in
obesity analysis acknowledged. 30] class A,B equity; class C,D racialdata; class E,F,G,H,I,J,K,L,M,N,O,P,AG health; class Q,R,S,T,U,V,W,X,Y,Z,AA,AB,AC,AD,AE,AF policy;


1.-Equity means meeting people where they are and providing solutions based on their different individual needs, not just equality.

2.-Two key projects: AIM AHEAD to increase diversity in AI for health, and Racial Data Tracker to highlight structural racism.

3.-Racial Data Tracker collects and disseminates data on racial disparities across topics to inform advocacy and policymaking. Launches this month.

4.-A person's neighborhood is a major determinant of their life expectancy. Wealth and race/ethnicity also intersect with neighborhood.

5.-Social determinants of health (conditions of birth, living, work) impact health more than clinical care. Includes socioeconomic and environmental factors.

6.-Life expectancy varies significantly by race in the US. Males have lower life expectancy than females across racial groups.

7.-Historically racist policies like 1930s redlining of neighborhoods continue to impact health disparities and segregation in those areas today.

8.-Machine learning analysis of satellite imagery can accurately estimate neighborhood-level obesity rates, outperforming other data like points of interest.

9.-New analysis looked at 164M street view images to quantify racial disparities in built environment and impact on health.

10.-Classified neighborhoods by racial majority and matched to 1930s redlining maps. Data covered 59K neighborhoods in many US cities.

11.-Focused analysis on built environment features known or potentially related to health like green space, housing, sidewalks, crosswalks, etc.

12.-Found strong lingering association between 1930s neighborhood classifications and modern racial composition, suggesting lasting impact of discriminatory policies.

13.-Most significant finding was poor quality/multi-family housing strongly mediated disparities in sleep and asthma. Housing is critical for health.

14.-Goal is to measure and highlight inequities so they can inform policy changes. Without measurement, disparities are ignored.

15.-Wealth and economic factors do intersect with housing and health, but discriminatory policies have also independently shaped neighborhood environments.

16.-Team is considering how to analyze potential future impacts of current policies, such as those implemented during COVID-19 pandemic response.

17.-Change over time is evident in some areas getting worse, others better depending on targeted interventions. Racial Data Tracker will show trends.

18.-Bias in data used to understand these issues is a real concern that has to be carefully considered and mitigated.

19.-Complex intersectional factors require targeted, community-engaged interventions, not one-size-fits-all policies. But some social determinants of health are consistent globally.

20.-Research itself can influence behavior and outcomes, so potential societal impacts need to be considered in study design and data sharing.

21.-Quantifying and visualizing disparities, even if already generally known, is important for motivating policymakers to make changes over time.

22.-Collaborated directly with public health practitioners on past projects to influence disease surveillance. Racial Data Tracker builds on COVID tracking.

23.-Beyond proving disparities exist, research should provide guidance on potential policy solutions in accessible ways to enable advocacy.

24.-Changing built environment, like adding green space, is an example of a policy solution the research could point to.

25.-Need to study other countries in their specific policy contexts; impacts likely vary and results shouldn't be overgeneralized.

26.-Examining policies with positive vs negative disparity outcomes could suggest how to design better corrective policies for the future.

27.-The 1930s redlining classifications were highly problematic and discriminatory; we should not return to such labels today.

28.-Less segregated neighborhoods tend to have better socioeconomic and health outcomes overall. Reversing impacts of racist policies is needed.

29.-Racial disparities in home ownership, loans, and appraisals persist today in cities like Boston as a legacy of historical discrimination.

30.-BMI as a metric for obesity can be problematic across diverse populations; an acknowledged limitation in some of the analyses presented.

Knowledge Vault built byDavid Vivancos 2024