TOTAL: {[ getCartTotalCost() | currencyFilter ]} Update cart for total shopping_basket Checkout

Privacy Perspectives | On Making Consumer Scoring More Fair and Transparent Related reading: US NTIA releases AI Accountability Policy Report

rss_feed

""

""

To score is human. Ranking people by grades and other performance numbers is as old as time itself. Consumer scores—numbers given to people to describe their characteristics, habits or predilections—are a modern day numeric shorthand that ranks, separates, sifts and otherwise categorizes people and also predicts their potential future actions.

In our modern sea of data, the resources to examine all relevant information regarding a decision are no longer feasible, so we use shortcuts. Consumer scores built using predictive analytics and fed by large datasets are the modern-day shortcuts to understanding individual consumer behavior. That’s why new and unregulated consumer scores abound. They are used widely in today’s world to predict consumers’ behavior, spending, health, fraud, profitability and much more. These scores rely on petabytes of information coming from newly available data streams, and some old ones.

Not all scores merit concern. Some scores, especially those that do not identify any individual consumers, are unobjectionable.

A major issue with scores today is that many of them are secret, with hundreds or even thousands of factors that are also kept secret. Secret scores can hide discrimination, unfairness and bias both in the score itself and in its use.

Some scores, though, do merit concern and attention. The scores I am most concerned about are those that incorporate as underlying factors attributes like health and medical information, sexual orientation, race, national origin, religion, gender and other factors that are widely considered protected or sensitive information.

I am also concerned about scores that target and predict the behavior of vulnerable groups. Protecting vulnerable people from any deleterious impacts of scoring is a necessity, not an option. People who suffer from a terminal or significant health or medical condition, who are financial vulnerable, who are either elderly or minors, people who are unemployed, all require the strongest protections. If a score identifies or targets these populations, the score factors, score models and score uses need to be unfailingly beneficial and have meaningful transparency, oversight and consumer controls.

A major issue with scores today is that many of them are secret, with hundreds or even thousands of factors that are also kept secret. Secret scores can hide discrimination, unfairness and bias both in the score itself and in its use. Trade secrets have a justifiable place, but secrecy that hides racism, denies due process, undermines privacy rights or prevents justice does not belong anywhere.

Limits on secrecy for some consumer scores may offer a middle ground.

Determining where new consumer protections are needed can only be done with meaningful transparency and cooperation from industry, and any scoring best practices need to be developed with full consumer participation in the dialogue. The credit score was kept secret from consumers for decades. We can all do better this time around and ensure new consumer scores are themselves scored for fairness, accuracy and transparency.

Knowing the elements but not the weights of a scoring system provides a partial degree of transparency and reassurance. Knowing that there is a scoring system and how and when it is used helps. Knowing the source and reliability of the information and scoring model used to make a score helps. Being able to challenge a score and correct the data on which it is based helps. Knowing that some types of information will not be used for scoring helps. Knowing that data collected for one purpose will not be used for another or in violation of law helps. Knowing that the person or entity running the scoring system is accountable in a meaningful way helps.

For various reasons, laws governing credit scores do not typically extend protection to the new consumer scores. We must build a useful continuum of what scores are unobjectionable and what scores require consumer protection. Among the scores I see as most problematic are aggregate credit scores, any unregulated consumer risk scores used for or impacting eligibility and any score containing as factors either information that would otherwise be covered under the Equal Credit Opportunity Act or that includes any health information as a factor.

Determining where new consumer protections are needed can only be done with meaningful transparency and cooperation from industry, and any scoring best practices need to be developed with full consumer participation in the dialogue. The credit score was kept secret from consumers for decades. We can all do better this time around and ensure new consumer scores are themselves scored for fairness, accuracy and transparency.

The World Privacy Forum is publishing a large report next week, a year in its research, regarding consumer scores. The report will be available Tuesday, March 25, at www.worldprivacyforum.org. I’ll also be speaking at the FTC Alternative Scoring workshop Wednesday, March 19, at 10 a.m.

photo credit: Free Grunge Textures - www.freestock.ca via photopin cc

Comments

If you want to comment on this post, you need to login.