Algorithms are a key factor in successful companies. Decision-making is based on data. Scientist Esteban Moro explains the advantages to businesses and users of employing mathematical methods to interpret behavior patterns.
Esteban Moro divides his time between Spain and the United States. This big data expert at Carlos III University and visiting professor at MIT examines the role of algorithms and their power in the society of today.
Is there always an algorithm behind successful companies today?
There is an algorithm behind today’s most successful companies. Take Google, for example, based on an algorithm known as PageRank, or the Uber platform’s algorithm. Big business is based around algorithms, and nowadays very few companies are not based on data analysis.
Are algorithms the solution to all the problems of today?
Algorithms help us go further in terms of decision-making. But you can’t just let them take the decisions automatically. You always have to have someone behind them to keep an eye on things.
Do algorithms have to be covered by legislation?
One of the positive aspects of decision-making based on data is that you can monitor the traceability of decisions. Algorithms must be open-format, accessible to the public and also to legislators, in order to ascertain whether there is any kind of discrimination. Every time a company takes a decision based on algorithms, the algorithm should not be private, or should at least provide detailed information to demonstrate there is no racial, sexual or ageist discrimination etc. involved, and ageist discrimination is quite frequent.
Will Google or Facebook eventually make information on their algorithms public?
There is no 100% knowledge of the Google search algorithm, but we do know how it works. The Facebook algorithms are not public, but considering what happened with the news feed algorithm [department of the company handling current affairs], I believe they should make a portion of the algorithm public to prevent fake news going viral, as this led to much controversy and could have compromised the outcome of the US presidential election.
Google and Facebook handle so many data that they have the capacity to produce an automatic algorithm-based solution to detect the reliability of news items. During the US elections, Facebook was directly accused of failing to implement any measures to counter the fake news propagated on its platform directly targeting Hillary Clinton and Obama. And 44% of Americans get their news through this media.
So, should they make them public?
The value of companies such as Facebook or Google lies in the data, not the algorithm. They will increasingly make their algorithms public because they know that the value lies in data. They also know that only they have the capacity to compile information from millions of customers.
The value of companies such as Facebook or Google lies in the data, not the algorithm
Is it easy to manipulate algorithms?
They have to be trained constantly. Consider, for example, credit card fraud detection algorithms: every time you swipe your credit card at a terminal, the operation goes to the bank to demonstrate that it’s actually you swiping the card, using an algorithm that ascertains that you usually go to that kind of shop, that this is the area you live in … in the milliseconds it takes between the swipe and the pronouncement of acceptance, there’s an algorithm there deciding whether it’s you or whether it’s fraud. But that algorithm can also fall into the hands of fraudsters, and so information has to be added constantly. At the end of the day, what the algorithm does is transform patterns of behavior into mathematics, and nothing is engraved in stone. If the pattern of behavior changes, you have to change the way in which you translate it.
What’s the best location for gathering data?
There are many sources of data – bank cards, applications, social networks, phones – but all of them have their own problems. The problem with Twitter, for example, is that we don’t know who’s behind an account. This is inferred, and if that is not possible there’s an algorithm for it.
What do users expect from algorithms?
As end users, we want them to make our lives easier. We want them to send us to the restaurant we like most, to meet the person we have most in common with … we expect them to help us take decisions that otherwise we would be unable to take because we can’t analyze so much information. In my case, for example, I can’t possibly know everything that’s going on around the world, and ultimately I use a newspaper, which is an algorithm with people who in turn have been trained with mental algorithms which select articles for me.
What does an algorithm have to have to make it successful?
Most algorithms today are based on the utilization of a large amount of data. The value of a company lies in its data.
What can they do for banks?
They’re already doing things for a bank such as BBVA. Finance companies have been producers and consumers of information and data for many years. It makes sense to use algorithms at a bank to analyze all the data it generates, and to offer better services to its customers. They can help people invest their money, or give them a better customer experience, or manage fraud. Data improve algorithms. For example, they enable people to benefit from credit even if they have no long-term financial background with the bank.