fbpx
Features Hub

Future Decision-Making Processes in the Data Science Era

Mon 10 May 2021 | Mohammad Mahdavi

What’s next for decision-making in the era of automation? Mohammad Mahdavi, Professor of Data Science at GISMA Business School, explores

Traditionally, decision-making was an intuition-based activity. Decision-makers have to rely on their own or their consults’ intuitions and experiences to run their business. Consider a supermarket business scenario. As an experienced retailer, you might notice you need to store boxes of chocolates before Christmas, eggs before Easter and beers before every weekend. You might also notice that customers usually buy certain items, such as butter and jam, together. So, you can put them close together in your shop to remind customers to buy both together.

Although the decision maker’s experience is a powerful source to improve the business revenue, it is limited. As human beings, we have a limited cognitive power to discover more subtle signals. For example, even a very experienced retailer might not notice that some types of customers with certain characteristics, such as age and profession, tend to buy certain kind of items. As a result, the business’s success is limited by the decision maker’s cognitive power.

However, data science has changed the game. Instead of relying on one person’s intuition and experience, data-driven approaches leverage huge amounts of data that a business has collected through time. These approaches extract knowledge and insights from historical data to support decision-makers in future decision-making processes. Looking back again to our previous example, in an online world, we can collect logs of customers’ previous purchases. Using this dataset, we can train machine learning-based approaches to automate the process of extracting the mentioned insights: Which items should be stored at which dates? Which items are likely to be sold together? Which customer is likely to buy which item?

Furthermore, by integrating the above dataset with datasets from other domains, such as weather forecasts, traffic jams and pandemic rates, we can automatically gain other more interesting insights: Is there any relation between the weather condition and the sold items? Does the traffic or pandemic affect our revenue?

Using these data-driven approaches, business managers can run their businesses more effectively and efficiently. An appealing characteristic of these approaches is that they are nearly automatic. Once you collect some data and train a model to predict something, you can apply it without any human effort. For example, once you train a recommendation system to recommend certain items to certain customers, they keep working without any needed human involvement. Although, we may need to monitor their performance from time to time and retrain them again with more updated data.

So far, data-driven approaches have been used mainly as so-called decision support systems. That is, they supported us to make decisions, such as recommending items. The final decision, whether to buy an item or not, has been mainly up to the human agents. However, this paradigm is shifting towards more automation.

It is not unrealistic to imagine machines taking more control in the future. Most of us might appreciate a computer learning our ‘shopping model’ and doing it on our behalf. Think about it; we reach our home at night and see the box of goods in front of our door, automatically purchased by our data-driven shopping consultant. Let us look at an escalated example; we are currently training autonomous cars!  This means, in near future, we may not even need to make any decision for driving our car.

Of course, this trend towards more and more automation in decision-making is both thrilling and frightening at the same time. More automation means less groundwork for human beings, which directly improves the quality of life by allowing them to focus on more enjoyable activities. On the other hand, it seems scary to allow machines to take control of every detailed decision of our life. It is an inevitable trend, but we can prepare and think about the necessary plans and regulations upfront.

Experts featured:

Mohammad Mahdavi

Professor of Data Science
GISMA Business School

Send us a correction Send us a news tip