Beyond open data: Open algorithms

No one questions the intense process of datification that we are living nowadays, through which data no only contribute to creating a more truthful image of the world, but also to shape and actively model it through the algorithms fed by data.

Algorithm can be defined as a set of invariable steps necessary to solve a problem or to carry out a task from some specific initial data. Algorithms have always been with us; however, its increasing use in decision-making processes, combined with a lack of transparency and control, can lead to deep social implications.

Algorithms are starting to replace people when making fundamental decisions in several spheres of our everyday lives. When we apply for a job, it is increasingly more frequent that the first selection is made by an algorithm, and the same happens when we apply for a car insurance. Currently the algorithms can decide whether we should go to jail or we can continue on bail, manipulate our emotions or even choose whether we should live or not in case of accident with our autonomous car. All of them are very important decisions that affect us directly. Nevertheless, at present we know very little or nothing about the data or criteria on which the algorithms are based to make such decisions.

As a society, we have been able to develop the tools to make the decisions, however we still have not found the formula to guarantee that these decisions are ethical and neutral, avoiding that our own prejudices as human beings are also transmitted to our algorithms giving rise to clear cases of discrimination. This issue becomes much more delicate when artificial intelligence and predictive analytics come into play, with new and more complex algorithms that not only execute step by step the orders of programmers, but learn from their own experience and make their own decisions.

That is why the algorithms are increasingly being questioned, for example in the current proliferation of false news in social media, a disease of our century according to Pope Francis and a threat to democracy in the words of the former President Obama. The concern is such that the European Commission is considering carrying out actions to make Internet companies share the details about the algorithms they use when presenting certain personalized information, such as news or online search results.

The options we have for more open and transparent algorithms are several, as discussed at the recent OGP Summit:

1. Minimizing the data collection.
2. Optimizing the algorithm transparency and openness.
3. Auditing algorithms.
4. Developing normatives and standards for the algorithm behaviour.
5. Developing legal frameworks to guarantee protection.

The discussion is already on the institutions’ table and the first steps to face the problem are being taken, so we expect to continue to talk about openness beyond open data in the near future.