Pendyala is an assistant professor of applied data science at San Jose State University and is a public voices fellow of The OpEdProject.
In a recent study, 71 percent of the government leaders said that using generative artificial intelligence in their operations will result in benefits that outweigh any potential risks. Maybe we should enshrine the use of data in our founding documents.
Article I, Section 1 of the U.S. Constitution states, “All legislative Powers herein granted shall be vested in a Congress of the United States, which shall consist of a Senate and House of Representatives.” I think it is now time to append that statement with: “supported by a body of data science tools and technologies.”
Decision-making in industries, including the restaurant business, is increasingly being driven by data. Even some investment funds are now entirely automated from end to end, using data-driven decision-making. But governments are still slow to adopt data-driven methods, even for the most widely impacting policies. In democracies, legislation is still predominantly driven by vote-bank politics and popular or partisan beliefs.
While data science is not yet devoid of biases and other ethical issues that the recent presidential executive order on AI addresses, the science is still mature enough for application in governance – with reasonable checks and balances provided under the U.S. Constitution.
Sign up for The Fulcrum newsletter
Last year, I submitted a grant proposal to investigate the safety of a legal right turn on red using machine learning. It was denied, even though further analysis will prove the efficacy of using of AI to investigate government programs.
For example, I asked ChatGPT: “Based on the data you have, can you tell if making private elementary school fees tax deductible will help society?” ChatGPT is trained mostly on data in the form of natural language statements available on the Internet and not on specific numbers. Still, it did a useful qualitative analysis of the impact by providing detailed and sensible arguments for and against the concept. Using quantitative data certainly is a tool for providing more specific metrics.
Data may also show how much time, energy, and effort spent in commute may have been saved, and accidents avoided, if the government incentivized employers and schools of a minimum size to run shuttles for commuters. More data can probably also quantify the money saved if the resulting climate disasters could have been avoided. We currently do not have these studies, hence we are having to rely on politics and opinions that aren’t supported by accurate data.
As AI advances, and becomes more commonplace, we need to open our mind to how it might reduce unsubstantiated electoral promises, abuse of public services, and government inefficiencies. Data will certainly help in determining if laws like California’s Proposition 47, which reduced certain low-level drug and property crimes from felonies to misdemeanors, did more bad than good to the community – not people or the popular vote.
As part of my talks at events and the class that I teach on Big Data, I explain how a Massachusetts governor’s own sensitive health information was exposed by a graduate student despite his assurances that the common man’s privacy is protected.
President Biden might have been able to avoid contributing to the rise of inflation if there was an additional check done by data-driven methods. Building prediction models using artificial intelligence based on a number of current and past features of the economy, analysts could have predicted how Biden’s proposals would affect inflation.
People’s opinions may not always be correct, but data analysis can expose latent characteristics of the situation described by the data.
