Tech giants can predict every consumer’s wants and needs using large datasets and machine learning algorithms. Google, Netflix and Amazon commercially apply the “black box” paradigm to machine learning, which fuels the recommended purchases, TV shows and films you see on Amazon and Netflix. The second approach to machine learning is the “causality” paradigm used at scientific institutions.
With the black box approach companies use ML algorithms to match your preferences with other customers and make recommendations based on data from your peer group. Depending on your feedback, it may adjust the peer group. The algorithm simply produces forecasts based on preferences, but doesn’t build a theory and doesn’t understand the cause-effect mechanism behind the preferences.
Researchers’ causality paradigm is fundamentally different to this. Scientists aim to derive a theory of how X causes Y, using machine learning to narrow the search for potential Xs. While accurate forecasts are always useful, the ultimate goal within the causality paradigm is to explain why things happen, and to refine and perfect theories.
Like tech giants, financial companies are investing in machine learning to modernise the industry. However, many in finance focus on the wrong kind of it. Motivated by profits, they’re more interested in black box ML instead of building financial theories.
Focusing on capitalising on recent ML successes and building financial black boxes will set them up for disappointment in the long run. Noisy data sets and dynamic finance systems make it impossible for financial black boxes to keep up and produce financial theories that give managers and investors the confidence to commit to a strategy.
See the full article here: