Can AI help governments clean out bureaucratic "Sludge"?

 Government services often entail a plethora of paperwork and processes that can be exasperating and time-consuming for citizens. Whether it's applying for a passport, filing taxes, or registering a business, chances are one has encountered some form of sludge.

Sludge is a term coined by Cass Sunstein, in his straightforward book, Sludge, a legal scholar and former administrator of the White House Office of Information and Regulatory Affairs, to describe unnecessarily effortful processes, bureaucratic procedures, and other barriers to desirable outcomes in government services.

Sunstein has a unique point of view on this issue, given his work at the White House on eliminating paperwork and unnecessary bureaucracy. He contends that sludge is not only a nuisance, but also a grave threat to human dignity, welfare, and democracy. He asserts that sludge can impair health, perpetuate poverty, obstruct democracy, and forfeit jobs, scholarships, and opportunities. He also cautions that some sludge is intentional and manipulative, designed by companies or governments to exploit or discourage people from making certain choices.

One example of sludge in government is the complex and lengthy process of applying for disability benefits in the US. According to a report by the Government Accountability Office, applicants have to fill out multiple forms, provide extensive medical and work history information, undergo medical examinations, and wait for months or even years to receive a decision. The report also found that many applicants face challenges in comprehending the eligibility criteria, obtaining the requisite documentation, and communicating with the Social Security Administration. As a consequence, some applicants give up or die before receiving their benefits.

According to Sunstein's estimate, Americans spent 9.78 billion hours on federal paperwork in 2015. That is equivalent to about 4.7 million full-time jobs. Imagine how much time, money, and energy could be saved if AI could help cut the sludge.

How can sludge be reduced or eliminated?

So how can sludge be reduced or eliminated in government services? Sunstein suggests that one way to achieve this is to conduct Sludge Audits, which are systematic evaluations of the costs and benefits of existing or proposed sludge. He also recommends that governments adopt ethical principles and guidelines for the design and use of public services. He argues that by reducing sludge, governments can enhance the quality of life and well-being of their citizens.

One example of sludge reduction in government is the simplification and automation of tax filing in some countries. According to a study by the World Bank, countries that have implemented electronic tax filing systems have reduced the time and cost of tax compliance for businesses and individuals. The study also found that electronic tax filing systems have improved tax administration efficiency, transparency, and revenue collection. Some countries, such as Estonia and Chile, have gone further by pre-filling tax returns with information from various sources, such as employers, banks, and other government agencies. This reduces the burden on taxpayers to provide or verify data, and increases the accuracy and completeness of tax returns.

Future Opportunities for AI in Cutting Sludge

AI technology is rapidly evolving, and its potential applications are manifold. Here are a few opportunities for further AI deployment:

• AI-assisted policy design: AI can analyze vast amounts of data to inform policy design, identifying areas of administrative burden and suggesting improvements.
• Smart contracts and blockchain: These technologies could automate complex procedures, such as contract execution or asset transfer, reducing the need for paperwork.
• Enhanced citizen engagement: AI could personalize government services, making them more accessible and less burdensome.

Key Takeaways:

• AI could play a significant role in policy design, contract execution, and citizen engagement.
• These technologies hold the potential to significantly reduce sludge.

Risks of using AI to reduce sludge

While the promise of AI is significant, there are also risks and challenges that must be considered:

• Error propagation and appeals: AI systems can make mistakes, and an increase in errors could lead to more paperwork and appeals.
• Calcifying biases: AI systems can replicate and amplify historical biases in data, potentially exacerbating inequalities.
• Complexity and inaccessibility: The complexity of AI systems could make them difficult for citizens to understand and use, potentially adding to the sludge rather than reducing it.
• Bias and discrimination: AI systems may produce unfair or inaccurate outcomes that affect people’s rights, opportunities, or well-being, based on their personal characteristics, such as gender, race, age, or disability. For example, AI systems that are used to screen job applicants, allocate benefits, or predict criminal recidivism may be biased by the data or algorithms they use.
• Transparency and accountability: AI systems may be complex or opaque, making it difficult to understand or explain how they work or why they make certain decisions. This may affect people’s trust, consent, or ability to challenge or appeal the decisions made by AI systems. For example, AI systems that are used to diagnose diseases, grade exams, or recommend policies may be based on hidden assumptions or logic that are not disclosed or justified.
• Privacy and security: AI systems may collect, process, or share large amounts of personal or sensitive data that may be vulnerable to misuse, abuse, or breach. This may affect people’s privacy, confidentiality, or identity. For example, AI systems that are used to monitor public spaces, track online behavior, or identify faces may infringe on people’s privacy or expose them to surveillance.
• Human dignity and autonomy: AI systems may affect people’s dignity, autonomy, or agency by influencing their choices, preferences, or emotions. This may affect people’s self-respect, self-determination, or well-being. For example, AI systems that are used to nudge behavior, manipulate opinions, or generate content may interfere with people’s freedom of thought, expression, or action.

Key Takeaways:

• AI systems may increase paperwork if they are prone to errors or are difficult to use.
• The risk of calcifying biases is a serious concern that needs to be addressed.

Finally, there is the risk that using AI to eliminate burden may actually make the process worse. For example, if an AI system makes errors or produces inaccurate results, this may result in increased paperwork due to appeals or other corrective measures.

Conclusion: Get Excited & Mind the AI hype

While AI can offer many opportunities and benefits for cutting sludge, it is important to be realistic and cautious about its limitations and challenges. AI is not a magic bullet that can solve all the problems of bureaucracy and governance. AI is still a nascent and evolving field that faces technical, ethical, social, and legal hurdles. AI is also dependent on the quality and availability of data, the robustness and validity of algorithms, the competence and responsibility of human actors, and the trust and acceptance of users.

Moreover, AI may not always be the best or most appropriate solution for cutting sludge. Sometimes, simple fixes or low-tech solutions may be more effective or efficient than complex or high-tech ones. For example, a recent article by Bloomberg highlighted how Code for America (CfA), a nonprofit organization that aims to improve government services through technology, used SMS messages to help low-income people access food stamps during the pandemic. The article showed how remarkably effective this simple fix was in reducing paperwork, wait times, and errors, while increasing enrollment rates and user satisfaction:

"A surprisingly small number of states use texting, which is cheap—usually less than a tenth of a cent apiece—and can include links to application and renewal websites as well as allow recipients to reply with questions. When CfA worked with Louisiana to send texts in 2019, renewals increased by 37% in SNAP and 67% in Medicaid. In Michigan, texting improved approval rates from 53% to 67% and reduced the approval period from 13 days to 10 days. 'This is something we can really vouch for with confidence...'"

The article also showed how these sludge problems—particularly when it comes to benefits—will require more than some lines of code: new people, new rules, and time, lots of time, to slowly but surely make progress. For sludge-heavy programs, the immediate benefit of AI to reduce sludge may be limited (or hyped up). However, the article also underscores that truly addressing sludge—especially in areas heavy with it, like benefits programs—will require more than technological fixes. It will need new people, new rules, and lots of time to make slow but sure progress. Thus, while AI may offer some immediate benefits, its potential to cut sludge may be limited—or at least, hyped up

There is no doubt that AI has made significant advances in recent years. However, it is important not to get caught up in the hype surrounding this technology. While trying to cut through the sludge with these seemingly transformative technologies, it is important to remember that AI is not a magic bullet. It requires careful design, oversight, accountability — and most of all, thoughtful people focused on the end goal and not the technology of the moment.