Predicting behaviour risk starts with AI, psychology, and data you're ignoring

In the wake of the 2008 financial crisis, banking regulators across the globe began to recognize the need to increase the resiliency of global financial systems. While the majority of regulatory reform has been focused on containing structural and quantitative risks, the crisis also laid bare the outsized role that corporate culture had, and continues to have, on bank behaviour, the banking system, bank customers, and broader society. Combating the cultural drivers of misbehaviour continues to be one of the most difficult challenges facing the industry today.

Now into its second decade, the Wells Fargo saga continues to unfold, with no end in sight. The collateral damage to-date includes 5,300 terminated employees, $600MM in lawsuits, and $12.5B in fines, yet the bank continues to struggle with poor judgement and bad behaviour. While the company’s c-suite has apologized for its misdeeds and claim that the bank’s culture has reformed, evidence indicates this is little more than lip service: According to interviews conducted by the New York Times in March 2019, employees claim they continue to be pressured to squeeze customers and witness colleagues breaking rules to meet aggressive sales targets.

The banking sector is by no means the only industry to face these challenges. An ever-expanding list of companies like Volkswagen, Uber, Facebook, Google, and Boeing have seen their internal cultural challenges become very public debacles that have hurt customers and shareholders, irreversibly tarnished their brands, with global societal repercussions whose impact will be felt for decades.

Shifting from a reactive to preventative approach

The Australian Prudential Regulation Authority’s 2019 report on the Commonwealth Bank of Australia highlighted the existence of a multitude of the non-financial risks relating to culture, and emphasized the necessity of an overall shift in philosophy toward banking culture, and specifically the need to move from reactive approach to one characterized by identifying the cultural early warning signals that have the potential to manifest into broader problems.

Most financial institutions employ traditional “reactive” approaches to monitoring employee behaviours through the use of technologies like User and Entity Behaviour Analysis systems, Trade Surveillance systems, and Security Information and Event Management systems. These “reactive systems” were not designed for prevention, rather they identify problematic behaviour only once they are occurring, and in doing so are not able to predict and prevent unwanted behaviour from happening in the first place. By the time problematic behaviour is identified by these systems, the damage has already been done.

For the most part, financial institutions' inability to anticipate unwanted behaviour has resulted in a reactive approach to managing conduct risk, where forensics plays an outsized role and an overwhelming volume of false positives consume significant resources that could be better deployed focused on prevention.

Hidden and unpredictable culture risk

While executives play a role in defining culture and the parameters of acceptable behaviour, this assumes that the board and executives have overwhelming control over employee behaviour. In reality, most organizations suffer from cultural risks that are near-impossible to see, and the highly hierarchical organizational structures that predominate in banking further exacerbate this problem.

What is likely the most difficult aspect of managing risk culture and predicting behaviour stems from the unexpected. While executives play a key role in establishing culture and the ultimate thresholds for acceptable behaviour, these thresholds are established and practiced within typical, expected, and foreseen operating environments. In times of change, strain, upheaval, crisis, and when internalities and externalities stress the operating environment in unpredictable ways, the defined culture and practiced parameters for acceptable behaviour may no longer be able to contain actual employee behaviour. This systematic failure makes it near-impossible to predict or prevent unwanted behaviour, and ultimately leaves banks and the banking system exposed to incalculable risk.

Prediction vs. insight engines

While artificial intelligence is being utilized to solve an ever-expanding number of previously unsolvable challenges, one can be forgiven for assuming the silver bullet is AI-based employee behaviour prediction. As humans, both our psychological state and our behaviour is influenced by a complex combination of internal and external factors that are highly unique to each and every one of us – our individual previous experiences and our individual current circumstances. While consciousness allows us to actively explore the world around us and make choices by imagining the anticipated consequences of our actions, externalities serve to both influence and limit our choices. For this reason, a model that can accurately anticipate behaviour without intimate knowledge of intangibles like our consciousness and life experiences remains in the domain of science fiction.

The black-box nature of AI systems also presents significant challenges to creating systems to identify employees who are at risk of problematic behaviour. As opaque algorithms make it near-impossible to explain why conclusions are made, employers will face significant challenges changing their problematic cultures when the contributing factors to the cultural problems can’t be isolated and identified. Legislative barriers are also on the horizon: should the Algorithmic Accountability Act of 2019 become law, it will serve to regulate automated decision-making, further restricting the use of black-box decision-making systems in combatting cultural and behaviour risk.

A preventative solution then, should not be one that employs opaque algorithms to predict what people are going to do, but rather one that realizes prevention by being explainable, and by virtue of being explainable can also be used as a basis to make changes to the organization that will sustainably reduce culture and behaviour risk.

Language psychology and the data you’re ignoring

While traditional NLP techniques focus on keywords, topics and themes - primarily conscious language, Receptiviti’s proprietary approach focuses on unconscious language coupled with data science, and machine learning, and is backed by a significant body of social psychology and language-based research into mental and cognitive health conducted over the past two decades, primarily by James Pennebaker, Chief Science Officer at Receptiviti.

The signals that emerge enable organizations to understand the psychological health of their workforce and its influence on employee behaviour through leading indicators like stress levels, risk taking propensities, the social dynamics that exist among employees, teams and groups within the business, as well as team cohesion, and informal influence hierarchies. Risk prevention is made possible with a real-time view of changes in workforce psychology and emerging outlier groups whose prevailing or trending psychological state puts them at risk of behaviour that is at odds with company norms.

A new role for risk leadership

In the post-financial crisis world, the definition of company culture has expanded to include risk culture. Traditional survey-based approaches that have been utilized by human resources for decades are now widely recognized as an inadequate mechanism for measuring risk culture and preventing unwanted behaviour. Preventing lapses like those seen in 2008 requires far more sophisticated tools and approaches.

While in years-past company culture was the domain of human resources departments, the responsibility now falls on executive leadership, boards, compliance, and risk departments who need to secure their own solutions to identify the cultural outliers that continue to be one of the biggest risks facing the global banking industry.

Introducing Mike Friedman, Receptiviti’s New Chief Analytics Officer and Head of Product

I’m very excited to announce that Mike Friedman has joined Receptiviti as Chief Analytics Officer and Head of Product. Mike was previously Managing Director, Data Science & Analytics at Scotiabank. For almost two decades years Mike has been building and managing large teams of data scientists and engineers to design, develop, and deploy advanced analytics and artificial intelligence based systems across the bank, including Scotia’s Corporate and Investment Bank, Global Transaction Bank, and Commercial Banking.

I was introduced to Mike in 2016 by one of our advisors, and we quickly developed a great working relationship. Mike’s theoretical and practical knowledge of artificial intelligence, and his experience designing and building highly complex analytics systems that simplify business problem solving was incredibly appealing. I knew early on that we wanted to bring him onboard. In addition to his professional experience, Mike holds degrees in Commerce and a Master’s in Management Analytics from Queen’s University. He is a Chartered Accountant in Canada, a Certified Public Accountant in the U.S., and an adjunct lecturer at Queen’s Smith School of Business.

Our customers hold us to an incredibly high ethical standard, and as we continue to scale it’s increasingly important for us to bring on established leaders like Mike who have deep expertise in ethical AI and are highly regarded both in industry and academia. There’s an elegance and empathetic nature to Mike’s data-driven approach that is quite rare, he’s relentless when it comes to simplicity and actionability, both of which are critically important when building systems that people actually want to use.

As we continue to grow our customer base and partnerships, expand our research and the capabilities of our platform, Mike’s experience will be a huge asset and driving force into 2019 and beyond. I’m delighted to welcome Mike to Receptiviti’s leadership team!

See the Betakit article for more information about Mike.