Two important and arguably interdependent revolutions are taking place in wealth management today–a focus on client best interest standards from both ethical and regulatory standpoints and a growing use of algorithms to personalize the investor experience and create compliance and operational efficiencies for dealers and advisors.
Algorithms are the backbone of many of our everyday processes, from the innocuous (online shopping) to the significant (economic forecasting). However, algorithms aren’t perfect. There are numerous and, at times, devastating cases of algorithms creating undesirable outcomes due to data flaws or unintentional biases built into the system. Such cases compel us to ask the questions: who is responsible for the decisions that algorithms make and who is accountable for any unintended outcomes?
Increased Reliance on Algorithms Calls for Increased Accountability
As the industry moves toward client best interest and we consider the potential risks of harnessing algorithms for complex wealth management processes, we realize the profound responsibility on the shoulders of wealth management firms to ensure that their algorithms are held to fiduciary standards. The question is: how can the duty of care that has traditionally been applied to advisors be applied to the technology they now employ?
Fiduciary Algorithms Follow These Core Principles
Writing for the MIT Technology Review, Nicholas Diakopoulos, from the University of Maryland, and Sorelle Friedler, from Haverford College, introduce principles exploring how to hold algorithms accountable. They argue that the following five principles should be considered during all stages of design, development, implementation and use in order for an algorithm to be a fiduciary (we, at Univeris, have added a sixth principle):
Responsibility — Systems are in place to collect feedback and to monitor, evaluate and improve the algorithm.
Explainability — The algorithm provides clear evidence of client best interest. For this to be possible, the algorithm and its data sources are explainable to all users and stakeholders in plain, non-technical language.
Auditability — Traceable, transparent and impeccable records are available for review by all users and stakeholders, as well as by third parties when needed.
Accuracy — Errors are flagged and resolved in the algorithm’s data source in order to improve the statistical confidence of the algorithm’s decision.
Fairness — The algorithm’s calculations are unbiased and its results, therefore, do not discriminate.
Anticipation — The algorithm “senses” the meaning and ramifications of the specific objective it has been set to resolve for the client.
Fiduciary Algorithms are a Key to Digitalization in Wealth Management
With CRM3 on the horizon, algorithmic accountability should be an integral part of the digitalization journey for wealth management firms.
In the positioning paper, Algorithmic Accountability in Wealth Management we provide an in-depth exploration of the principles of fiduciary algorithms and how they can be applied to wealth management.