South Africa

Machine learning and artificial intelligence come to corrupt officials

South Africa has a major problem with corruption in government supply chains. The most striking example of the recent Covid-19 pandemic was the looting of funds, particularly the purchase of personal protective equipment at the Gauteng Health Department. Mark Heywood right he claimed in Napi Maverick that if we do not introduce security in the punishment of corrupt public officials, we will lose the fight against corruption.

The looting and riots in July have taught us that these events are affecting our daily lives. They are causing job losses and rising food prices and are particularly severe in the youth sector. Even the Cosatu union federation acknowledged In 2017, corruption will cost at least 27 billion rupees and 76,000 jobs a year. That was before the epidemic.

  • Now imagine a scenario in which an accounting officer in the Department of Health can accurately predict the likelihood of ineffective and wasteful spending and take action to prevent it?
  • Think for a moment about the power of a transformer that can predict the likelihood of xenophobic attacks or unrest, such as the now infamous one. The riots of July 2021to be completely avoided?
  • What if an entire government supply chain can be treated as a blockchain by a distributed general ledger without any civil servants involved?

Business areas, including the banking and insurance sectors, have been making such forecasts for some time using data science tools. By applying machine learning to predict the risks of their own business models, they predict the likelihood that a customer will default on a loan or initiate an insurance claim. Have you ever wondered why the bank doesn’t want to lend? There is an algorithm in the background!

Data science is an emerging field of research that is usually associated with buzzwords such as big data, machine learning, and artificial intelligence (AI). The roots of all these terms are in classical statistics. Learning statistics is quite simply learning from data.

This was made possible by two conspiracy realities: the cost of storing data has decreased over the years, and computing power has increased. exponentially. This means that patterns and relationships can be found in very large data sets (hence the term big data).

One way to understand this capability is that if the data is too large for an Excel spreadsheet or a central processing unit to handle, it could be a data science task.

What if we brought data science and good governance into a chat room?

I believe that such an approach will bring huge benefits in terms of good governance, evidence-based policy-making and the fight against corruption. Full Disclosure: I wear more than one hat.

As the legislature of the Gauteng Provincial Legislature and a member of the Standing Committee on Public Accounts (Scopa), I often hear well-founded complaints about the method of ex-post oversight. The sectoral supervisory model adopted by the South African legislative sector is a retrospective tool. Supervision typically takes place months after irregular public spending. The committee’s recommendations do not deter corrupt officials.

As a social scientist and novice coder (Python is good fun!), This makes me think about the catalytic mechanism of an artificial neural network or a decision tree regressor in the fight against corruption.

However, the specific model is not as important as some reality checks.

  1. In the web-native world of data science, anyone can write code, teach a computer model, and release it on real data. This is indeed encouraged by the fact that free and open source resources are now ubiquitous. Urban and Pineda properly problematic this is a wealth of free information: Most are not rigorous enough to warrant a thorough investigation, and few resources, if any, are directed specifically at the decision-maker.

A simple Google search will reveal thousands of short-form resources in the form of “How To” video tutorials, articles, listings, and blogs, each covering a specific part of countless elements of data science. Topics like “Preparing Data with Pandas” or “How to Choose Features and Answers” ​​or “How to Tell if My Choice Algorithm Works” all give tempting glances at problem solving.

The world of data science is rarely systematically expanded, referenced, and peer reviewed specifically for decision makers, legislators, and government officials. Yet, on the periphery of applied policy-making, most officials are aware of concepts such as big data, machine learning, and AI. These concepts need to be clarified before they can be formally introduced in the field of governance.

  1. The second question is whether the data exists. Let me explain. Data on government performance are ubiquitous and abundant. In my case, Scopa members are constantly flooded with data. Data sources include the reports of the Ministry Committee, the Office of the Court of Auditors, the Civil Service Committee, the Finance and Taxation Committee, the Special Investigation Unit, the internal audit reports, the quarterly reports of the departments, and the list can be continued.

However, amid all this information, I very much doubt that there is a data set that is ready for machine learning. If anyone is reading this, you want to refute my claim, I would welcome such a development. This will save me months of research!

  1. The third question is reproducibility. If my team and I are building a machine learning model that performs well on unseen data, we need to share! It should be common practice to make not only the data sets but also the actual code available as a standard part of the research. This is because sometimes the algorithms don’t work, machine learning models deteriorate over time, or we make bad business decisions from the data. In such cases, we can all learn from our failures as well as from our victories.

For the policy maker, machine learning can become a tool to help prove or disprove our intuitions about the problem we want to solve.

Here are some suggestions for stopping corruption with a data-driven approach:

First, governments need to become a data-driven learning organization. Much more research and experimentation is needed to make this happen. Government officials and area experts are vital to the success of a business because they draw from the data the vital conclusions on which, say, decisions to eradicate corruption depend.

Second, we need to find the place of data science in the policy life cycle. Ideally in multiple locations. Statistical models can help policy makers move from inputs and outputs to results and impacts. This depends on the computational efficiency available and the specific problem the machine is trying to learn about, but machine learning can be beneficial at all stages of the policy lifecycle.

The next step is to bring together the data scientist, statistician, and domain expert. Governance is an extremely complex task. Every moment of the day, thousands and thousands of financial transactions take place through municipal, provincial and national budget lines, which often involve amazing amounts. All this takes place in an environment governed by a myriad of complex laws and regulations. Officials often underestimate their territorial knowledge and do not receive recognition for their ability to fuse this complexity. The current consensus the fact that the best problem-solving teams include data scientists, statisticians, and experts in the field. All three have a separate role to play in making informed governance decisions.

The opportunities for machine learning to fight corruption are very exciting. But it still requires humility from the data scientist, an understanding of the theory, and a willingness on the part of the government official to pass on his knowledge freely. And finally, no model can predict with 100% accuracy. We need to be honest about this when we offer solutions to both legislators and governments.



Machine learning and artificial intelligence come to corrupt officials

Source link Machine learning and artificial intelligence come to corrupt officials

Back to top button