Data Science vs. humans: who to trust in decision making?

o does this permeability to decisions made by artificial intelligence mean that soon we will be able to let algorithms make the decisions that matter in our organisations and, more broadly, in our society? As the technology itself evolves and becomes more effective at detecting patterns and preferences, should we let it point us in the right direction? Will data science algorithms make human intuition obsolete? Two main reasons may lead to a negative answer to these questions:

1. The application of data science to human behaviour needs human subjectivity

Typically, artificial intelligence algorithms analyse large volumes of data in order to find patterns associated with a particular group or class. For example, we can train an algorithm to distinguish images of dogs and cats by giving it enough images of each of the categories so that it learns the characteristics that are most associated with each of the animals.

Photo

It has been shown that these algorithms can indeed learn these patterns efficiently and accurately, not only in tasks such as this, but also in such important problems as predicting road accidents [2] or identifying people at risk of suicide [3][4].

It is precisely in sensitive human problems like the latter that it becomes clearer that algorithms cannot have all the answers. On the one hand, there may be people who do not fit the patterns detected, whose identification requires a degree of empathy and sensitivity that only another human being can offer. On the other hand, while an algorithm can suggest actions to follow, the validation and application of those actions will have to be done by people, in relationship with other people, with all the subjectivity inherent in human relationships.

And this idea does not only apply to these more extreme cases: for example, managing the relationship with a dissatisfied customer may also require a degree of sensitivity that complements or even contradicts the action suggested by the algorithm.

However, this does not mean that data science algorithms should be discarded. On the contrary: they can be a fundamental tool to destroy preconceptions, suggest new paths, manage clients and identify cases invisible to the human "eye". Just emphasise that we should not hand over all the decision-making power to an algorithm, being fundamental to leave room for more subjective decision factors.

2. The process of developing data science algorithms integrates human subjectivity

Although we sometimes tend to look at data science algorithms as something totally objective, it is important to recognise that they also have a subjective component that arises from their very development. In fact, less objective human aspects are also an integral part of the process of developing data science models.

The quality of any data science model starts by being determined by the quality of the question it asks. To give a very simple example, asking who are the customers that will join a particular campaign is not the same as asking which campaign will make which customers more satisfied. Similarly, the decision of what data to collect, and the format of that data also stems from subjective decisions. "Garbagein, garbage out: an algorithm can only deliver good outputs to the extent that it is fed with complete, adequate and unbiased data, and it is up to people to ensure this.

In conclusion...

Data science applications will be increasingly present in our day-to-day lives, informing and guiding decision-making in a wide variety of fields. These applications can make a contribution that goes beyond the human capacity to analyse and answer complex questions. However, it is important not only to understand the potential of data science, but also its limitations - recognising the need to include a subjectivity that (at least for now) only humans are capable of.