Stanford Talk: Inequality in Healthcare, AI & Data Science to Reduce Inequality - Improve Healthcare | Summary and Q&A

TL;DR

Using AI, this analysis explores the social implications of AI, studies inequality in policing, pain, and COVID-19, and highlights the importance of fine-grained mobility data for understanding disparities.

Key Insights

• 👮 There are significant racial disparities in police traffic stops, with AI being used to study racial discrimination and identify potential bias in policing practices.
• 📤 Overlooked physical features in knee x-rays contribute to disparities in pain levels among different racial and socioeconomic groups, highlighting the importance of using AI to improve the accuracy of severity assessments.
• ☠️ Fine-grained mobility data is crucial for understanding disease transmission and predicting disparities in COVID-19 infection rates based on socioeconomic status.
• ❓ AI has the potential to both increase and reduce disparities in various domains, depending on how it is applied and the diversity of the data used.

Transcript

Read and summarize the transcript of this video on Glasp Reader (beta).

Questions & Answers

Q: How was the data on police traffic stops obtained for the analysis of racial discrimination?

The data was collected through data requests made by journalists to over 150 police departments over five years. The dataset includes information on millions of traffic stops, such as the race of the drivers and the reason for the stop.

Q: How did the researchers assess racial discrimination in police traffic stops?

The researchers analyzed the data to determine if racial disparities existed in who gets stopped by the police, who gets searched after a stop, and how policy changes affect these factors. They also used statistical tests, such as outcome tests, to determine if black and Hispanic drivers were searched more often and had lower hit rates compared to white drivers.

Q: How did AI help in studying inequality in pain related to knee osteoarthritis?

AI was used to analyze knee x-rays and identify physical features that were not captured by traditional severity scores. By training a convolutional neural network, the researchers could predict pain levels more accurately and determine if disparities in pain between racial and socioeconomic groups could be explained by these overlooked features.

Q: How did AI estimate fine-grained mobility networks for modeling disease transmission?

Cell phone mobility data from Safegraph, which provides information on visits from neighborhoods to places, was used. By applying an iterative proportional fitting algorithm, the researchers estimated the true mobility networks from the noisy data, allowing for the modeling of disease transmission at a fine-grained level.

Summary & Key Takeaways

• Emma Pearson, an expert in AI fairness, discusses using AI to study inequality, focusing on policing, pain, and COVID-19.

• In the first part of the talk, Emma explains how AI is used to study inequality in policing by analyzing traffic stop data and uncovering potential racial discrimination.

• In the second part, she discusses using AI to study inequality in pain, specifically knee osteoarthritis, and how overlooked physical features in knee x-rays contribute to disparities in pain levels.

• Lastly, Emma showcases how AI can estimate fine-grained mobility networks and model disease transmission, revealing disparities in COVID-19 infection rates based on socioeconomic status.