AI in public policy: Pros and cons

Occurrences in the public sector impact everyone in the given society — from the military, educational institutions, law enforcement, and so on. Public policy enforced by local and national government policymakers plays a definitive role in how good, or bad the state of the public sector will be. More and more people are asking, “Can AI pave the path for better public policy?” these days. Believe it or not, there are many applications of AI in current public policy that provide some degree of change. With that in mind, it’s important to understand the current and future role of AI in public policy, the benefits and challenges it proposes, and how state governments can shape public policy with AI in a way that benefits society at large.

Regarding that, the main points we’ll touch upon today are:

• Connecting AI And public policy
• Can AI aid policymaking?
• Advantages and disadvantages
• Predictions for the future
• Key takeaways

Connecting AI and public policy

How much decision-making power can be bestowed to AI? To understand that, we need to acknowledge that the policymaking cycle is long and involves steps such as development, adoption, implementation, and evaluation across dozens of different sectors. AI is not entirely far from public policy and the government in the present day, as we’ll see in the examples below. However, can we expect to trust AI with more critical decisions, such as deciding whether or not a prisoner gets bail or which student is more deserving of a scholarship depending on their probability of success? The answer to the question is a little less simple. AI offers a data-driven solution to the public sector, but what about the human component? There are matters which rely on human intuition and judgment to achieve the best outcome. One thing we can do is accept that there isn’t a one-shoe-fits-all approach to fully integrating AI in public policy at the moment, and we’ll need a unique approach to each use case.

Can AI aid policymaking?

Enacting a policy is not an easy game: every decision has to be backed with substantial data to be considered valid, and that’s where your system may crack. Quite often, the pain points in policymaking stem from misinterpreted data, communication errors, or evidence that’s disintegrated into the decision-making process. AI in policymaking can cover these cases, helping decision-makers derive more value through the following:

• Pattern detection: AI can assist in drawing insights from large datasets and identifying recurring patterns in the quickest possible way. An example is using AI to find common properties to provide optimization solutions and devise interpretable policies.
• Forecasting policy: Forecasting is an evidence-based process. Putting the risk of data bias aside, responsible AI can be a great asset to bring OCR-based forecasts as dependable testimony into policymaking.
• Evaluating policies: To evaluate the efficacy or the impact of a policy and detect whether it meets its original objective, one has to track the behavior it provokes among the target group. That assumes the collection and analysis of enormous records of data, which gets down to our initial point of data collection and analysis through AI.

Advantages and disadvantages

Before AI policymaking can expand to have a greater role in the public sector across dozens of countries, it’s valuable to weigh the benefits and limitations it currently proposes:

Benefits

• Increased security — Better security anywhere from schools to the streets and national borders is a great advantage of AI that is partially exercised now. Facial recognition on border crossings and license plate detection are only a couple of uses of AI in law enforcement that establish local safety. If that would one day expand to smart detection systems being set up on streets and in stores, law enforcement could track criminal activity efficiently and mitigate it altogether by easily locating criminals.
• Anomaly detection & smart predictions — The opportunity to automate pattern detection instead of executing it manually proposes significant advantages for public policy. As we touched upon earlier, making data-driven predictions provides an edge that is more difficult to execute with manpower. With the capability to detect anomalies in data and make intelligent predictions, AI can be employed in all areas of public policy from the top down. The models can be used to find anomalies in financial records with the aim of mitigating financial fraud.
• Quicker processing times — There’s no doubt that a lot of the operations and processes that need to be carried out in any sphere of the public sector are time-consuming and laborious to do by hand. AI streamlines many of those processes, decreasing the margin of human error and revoking people of tedious work to allow them to concentrate on things that require their immediate attention instead.
• Accuracy — AI policymaking will heavily rely on factual data, which in its turn will promote a data-driven approach to any matter that is brought up, whether it be taking attendance of students in a classroom to calculating the probability of criminal activity threatening a specific area.

Challenges

• Privacy concerns — Considerable AI policymaking will leave people concerned about the invasion of their privacy more than ever. With a public sector fully emerged in everything AI has to offer, people will be “watched” everywhere: from schools to the streets, stores, and the workplace. All of that with the efforts to reduce criminal activity, implement smart attendance systems in schools, roll out employee-less smart stores, and much more. That will come at the cost of the government having access to immense data on residents' whereabouts. Will the tradeoff be worth it? It’s still a topic of controversy at the moment.
• Data bias — Avoiding data bias is another one of the public policy challenges that will need to be overcome before AI applications can be deployed to do the decision-making for public policy. It’s important to keep in mind that both machine learning and deep learning models need to be provided with training data before the model can fully function later on to make educated implications from that data. But how do we create a model that doesn’t discriminate by gender, race, age, and other characteristics and only concentrates on establishing a fair and well-off society via effective decision-making? It’s difficult to imagine right now. Keep in mind that if we broaden the range of the data to include more information, we may end up with an even more skewed outcome than initially.

Predictions for the future

Taking all of this into consideration, we can make educated predictions about how AI and public policy can become even more intertwined in the near future across dozens of countries throughout the world.

• Best of both worlds — Instead of leaving major policymaking decisions down to AI, they will be used with human supervision. That means if an AI model is programmed to, for example, analyze and predict the likelihood of a prisoner causing misdemeanors once they are released from jail, a judge will review the proposed data and then make a decision with it in mind, as opposed to leaving the full decision on AI prediction. This will temporarily aid in the challenge related to liability since the judge will be fully responsible for the decision and, therefore, held accountable instead of a machine.
• Skyrocket economy — It is predicted that AI will only become more of a cost-effective investment for anyone from the regular resident to the investor and the state economy as a whole. For example, the driverless vehicle industry alone “could save the United States $1.3 trillion in annual costs, or 8 percent of annual GDP, and $5.6 trillion globally once those technologies have fully penetrated,” as mentioned in a report from George Mason University. With the full emergence of autonomous vehicles alone, we can anticipate the average household fuel bill to drop, less time wasted on commutes, and more productivity instead, which will yield a greater return on investment.
• Governmental support will rise — More and more governments are allocating substantial budgets for AI, and that is suspected to grow over the years. In 2021 alone, the US government allocated more than 6 billion dollars to AI research and development, where a significant portion was granted to the US military for AI integration.

Key takeaways

It’s safe to say that AI has a role in the future of policymaking — in one form or another. While there are challenges related to heavily relying on AI, such as biases and discrepancies in data and the question of liability, there are dozens of advantages that we have to gain when utilizing AI policymaking in an effective manner. At the moment, the best path is to utilize AI in public policy but not withdraw human interference from the process entirely in order to fill in the gaps that AI can miss. Undoubtedly, we can learn how to apply AI to public policy outcomes to benefit society as a whole with minimal drawbacks in due time. One thing is for sure, all of the current predictions point toward a future with more advanced AI that encompasses all spheres of the public sector.