Do Data Scientists need a Bodyguard?

During the last decade corporations around the globe have been developing Data Scientist teams using different governing structures. From centralized units, to business functions. From internal units (costs centres) to federated with direct access to the market (revenue centres). Some of them focused on the technology set-up for scalability others more on the robustness of their analytical statements.

Whatever the starting point of all them they will eventually converge to a very fundamental challenge, Trust. Sorry for the spoiler, but yes, no matter how profitable and accurate your models will be, no matter how scalable your code is done, trust will be there waiting for you.

Organizations are represented by groups of individuals with their cognitive biasesat the very deep of them impacting on trust development, and therefore driving the direction of key decisions.

The development of behavioural economics has been remarkable in the last 20 years. However, their application has not been fully scaled into society.

So, what is need it?

We basically need environments that promotes openness, humbleness, reflection, and experimentation. A good starting point is to increase the organization awareness and maturity about two topics:

1. Scientific Method: from problem formulation, to hypothesis and analysis. Understanding statistics is key for that matter. Developing leaders accountable for most probable value and not black swan thinkers.

2. Cognitive biases: training intervention focused on bias reduction have a positive impact, at least is what was found in this study using games to target six different cognitive biases:

• Bias blind spot: seeing yourself as less susceptible to biases than other people

• Confirmation bias: collecting and evaluating evidence that confirms the theory you are testing

• Fundamental attribution error: unduly attributing someone’s behaviour to enduring aspects of that person’s disposition rather than to the circumstance in which the person was placed

• Anchoring: relying too heavily on the first piece of information considered when making a judgment

• Projection: assuming that other people think the same way we do

• Representativeness: relying on some simple and often misleading rules when estimating the probability of uncertain events

Playing the game reduced the three biases by about 46% immediately and 35% over the long term. Watching the video reduced the three biases by about 19% immediately and 20% over the long term.

Of course, these two are long term actions that will take time and resilience to have a significant impact.

Looking for quick wins?

If trust is about the human element, getting conscious about cognitive bias is a long path, what about using them for good?

Nudge Management

Developing a culture that supports thinking first vs acting first through nudgingcan have a short term impact. Remember that nudging is the creation of positive incentives for acting rationally.

In business nudging is pretty new and has been mainly applied in contexts like HR (i.e: Salary Inequality at Google, Savings for Retirement). In this experiment Virgin Atlantic used nudging to increase fuel efficiency of their pilots. But what about using nudging to minimize bias on key business decisions?

Influencers

Over last 5 years we’ve seen the boost of influencers through social media, creators of content sharing their personal stories. But if we scratch on the surface of the term, we will find a more classical reference, Opinion Leadership: “is leadership by an active media user who interprets the meaning of media messages or content for lower-end media users”.

Opinion Leaders counts with a group of supporters (lower-end media users) who follow their lead, ergo they trust them. Therefore, to stablish a good network of influencers can be a quick way to enable evidence-based decisions.

In the meantime, bodyguards

How many times the results of Data Science projects are compared with business believes, or outliers? I hope not many. Well typically, these are the perfect context for emotional decision making, delaying the application of insight or the deployment of Randomized Control Trials. This also can affect your team moral which could be even more problematic.

Having top stakeholder support is an accelerator for change but this should be tackled in parallel with long term actions. But please do not undermine our cognitive biases.

As the quote attributed to Albert Einstein says: Two things are infinite, the universe and human stupidity; and I’m not sure about the universe!

Further reading

Great article from Beck Hall, How can organizations build trust with their employees?.

My beloved Podcast Freakconomics with the chapter “Trust me” and of course The Hidden Brain The Halo Effect and The Double Standard

Nudge theory: the psychology and ethics of persuasion – The guardian.

Nudge management: applying behavioural science to increase knowledge worker productivity