In the use of data, the State has objectives that contradict each other

In the use of data, the State has objectives that contradict each other

In economy, most of the news has to do with statistics: what will happen to the poverty rate with accelerating inflation, or how the labor market will move if activity slows down. Even more so, in an election year, with the measurements of the politicians’ surveys. In an interview with Ámbito, economist Walter Sosa Escudero, specialized in econometrics and statistics applied to social issues, analyzes the dilemmas behind the numbers and how the use of big data could impact the national accounts system.

In addition, Sosa Escudero, a professor at the University of San Andrés and a CONICET researcher, enters into the debate on artificial intelligence and GPT chat and tells where he stands between the positions that propose that he is going to replace workers and those that establish that he is going to shorten working hours. Finally, she anticipates that she is working on a new book about how predictions work, to explain why it is so difficult to know how much the dollar will be or who will win the election.

Journalist: With the book “Big Data, a brief manual to learn about the data science that invaded our lives” you were one of the pioneers in writing on this topic, which some described as temporary. Today, 4 years later, do you think the State uses Big Data? In a State so questioned by society, could you use this tool to be more efficient?

Walter Sosa Escudero: There is a problem that the State has, like any complex organization, is that it has objectives that contradict each other. There are tons of things you could do for efficiency, like accessing individual opinion data, but for transparency that may be a limit. The case of a supermarket in the United States that began to design an algorithm to predict whether a girl was pregnant was famous, and if she detected it, she began to send her coupons with baby items. In an effort to sell more, the company cares little about the political costs of invading people’s lives. Now, the State has to ensure that right. Another example: the government of Norway, an aspirational country from many points of view, made public all the income of the population in pursuit of transparency. Within a month, everyone was talking about it, there were episodes of bullying, criticism of the lack of privacy, and they ended up backing down. So, this conflict of interest in companies is less than in the State.

Q.: Is that why the private sector is faster than the State?

WSE: There are the classic limits that bureaucracy has, but also the political cost of invading privacy. That is why I see it as natural that the State always goes a little further back, it has much more complex objectives than that of a company, which is often to generate economic revenue. We have made the word bureaucracy a pejorative meaning, but many times behind it is prudence. But that’s not an excuse, obviously, to slow down and incorporate Big Data. There is a point of prudence, but also a lot of space to work. Long-term plans and investments are needed to incorporate technology so that the advantages and not the disadvantages appear.

Q.: From time to time there is talk of modifying INDEC and making it an autonomous body, but the idea of ​​incorporating the use of Big Data into it is not discussed. What do you think?

WSE: INDEC has historically protected a set of very basic statistics, such as the measurement of unemployment, poverty, or domestic product, to name a few. What has happened in the last 20 years is that other sources of information have appeared. A concrete example is inflation, which is measured with a survey system. But now it turns out that with two clicks we can have information from thousands of online sites. I think that at some point the traditional statistical system of accounts should coexist with this alternative system that has to do with massive data. Unfortunately it is still a bit premature, not only in Argentina, but throughout the world. Because social statistics must be accompanied by a credibility service. It is one thing to give credibility to a scientific survey designed with a sampling frame and another thing to trust information from a company that may have short-term intentions to alter the information, or by an advertising campaign. A lot of institutionality is needed to provide Big Data data with that credibility that official statistics already have. All over the world there are doubts about how these two systems should coexist.

Q.: What could be the use of Big data in INDEC?

WSE: For example, if you want to know how poverty has evolved from the 1990s to now in urban agglomerations, you could make a model. Now, if you want to know how rural poverty evolved in the Pampas region, there is no data, you have certain information at certain times, in a country with a strong rural structure. Or if you wanted to know how the quality of employment in families living in marginal neighborhoods evolves, it is not known. While systematically looking at how the use of cell phones or certain interactions on social networks work, you could see if these groups are integrated into society. The information is urban, more systematization is needed, there are many places that are crucial for measurements. People have behaviors that used to be a black box, now you can see the decision map.

Q: Since we are in an election year, are we going to hear again that “it is useless”?

WSE: There’s something funny about the paradox of polls, which is that they perform worst when you need them most. One of the most predictable elections was the first for Barack Obama, because he was going to win wildly. When one is going to win or lose the surveys work very well. It is a very good job, but it is when it is least needed. When you need the stat the most is when things are pretty close. In short, when we face a polarized society, very divided, it is precisely when the polls work the worst. We should expect in the election year, that there is a kind of intelligence where the data provided by big data, social networks, is used to see if one can improve the performance of the polls. I don’t think the survey is useless, mostly because I don’t think people are stupid. So, if people use surveys, it’s because they give some information.

Q: To close, there remains the question on the subject of the moment: artificial intelligence and GPT Chat. Just as when Big Data arrived, some told you that it was a fad, how do you see this phenomenon?

The first thing that stands out is that incredible successes coexist with calamitous mistakes. But what really strikes me is how quickly anyone learns to detect where it works and where it doesn’t. My prediction is that for relatively minor tasks it will work very well. The problem that many people have with technology, with chatgpt, with big data, with the Internet at the time, is that they want to have binary positions of the type that will solve all problems or it is crap. If I ask him to write me a letter he will do it perfectly, but if he tells me ten measures to lower inflation he will say a string of nonsense. So, ultimately, I have to learn to see what it works for. The learning of the people is going to be very fast and the learning of the algorithm too. So I am optimistic, as long as we realize where the advantages are, where the disadvantages are. With the Internet, many people said that it was to look for pornography and suddenly the trade was going that way. With genetics they said that it was going to be used to build monsters and it turns out that it is used to cure diseases. What surprises me. It’s not so much how fast the algorithm learns, but how fast people learn to interact with the algorithm.

Q: Related to employment, very different positions are read, from the fact that it is going to replace workers to that it is going to make everything simpler and we are going to work fewer hours. Where do you stand between the pessimists and the optimists?

WSE: Goes to replace shapes. I dedicate myself to teaching and research. I have to retrain myself to see where I am irreplaceable to complement the algorithm. The role of the teacher passing on the information has its days numbered. I am going to have to focus my teaching role on discussing, giving examples, motivating, evaluating or giving feedback. What worries me is not so much if it is going to be replaced by robots, but how we are going to adapt to the changes. So, it definitely questions us about what we’re doing. The ultimate point is to think how much of what we do is automatic and replaceable and how much has to do with creativity. Chat has a tremendous advantage in generating content and humans in evaluating it. The interaction between the two things is what we still need.

Q: Last question, are you working on a new book?

WSE: Yes, about how we predict. And I am messing with all the forecasts, economic, meteorological, cultural and political. I’m interested in telling the backstage of how forecasts are made, why some work, why others don’t. Why are certain elections easy to predict and others not? Why did the short-term weather improve a lot? Why is it very difficult to know how much the dollar is going to be worth, or why in Argentina-France we were bolted to the couch without knowing what was going to happen. The book will focus on why some walk and others don’t.

Source: Ambito

Leave a Reply

Your email address will not be published. Required fields are marked *

Latest Posts

Two dead in Kiev airstrike

Two dead in Kiev airstrike

Image: SERGEI SUPINSKY (APA/AFP/SERGEI SUPINSKY) Two children were among the fatalities of the night attack with ground-based cruise missiles, said Mayor Vitali Klitschko and the