Week 2 Notes - Algorithmic Decision Making & The Census
Key Concepts Learned
Algorithm: a set of computational instruction for solving a problem or completing a task, especially in policy analysis and decision making
Algorithmic decision making in government contains input and output
Algorithmic decision making is especially appealing because it promises: Efficiency - process more cases faster; Consistency - same rules applied to everyone; Objectivity - removes human bias; Cost savings - fewer staff needed
U.S. Census data is the official collection of demographic, housing, and economic information about the population, conducted by the U.S. Census Bureau
Census data is used to allocate political representation, distribute government funding, plan infrastructure and public services, track social and economic change, and support academic and business research
Census data matters because it provides the foundation for fair public policy, ensures that resources are distributed where they are most needed, enables governments to understand communities, and supports designing fair algorithms and monitoring neighborhood change
Compared to census data, ACS data has more detailed information and will be more often to use
Large MOE relative to estimate = less reliable & Small MOE relative to estimate = more reliable
Coding Techniques
- We can use R packages to access census data directly
- Use get_acs() Function
Questions & Challenges
- How to find the proper ACS data and right variables for my research questions
Connections to Policy
- In the context of Public Sector, data collection has a long history with different types of data, and there also have some new data and transforming to prediction
- Algorithm always has problems in real world implications (e.g. data bias, human choice participation)
Reflection
- I assume the data is accurate, representative, and relevant to my research question
- People or groups underrepresented in the census might be excluded from my analysis
- My findings could reinforce stereotypes, justify biased policies, or ignore local differences
- I want policymakers to understand that my results depend on specific assumptions, data limitations, and interpretation choices, so they should be applied cautiously