December 21, 2022 | By Anik Bose, General Partner at BGV
Anik Bose shares his perspective on the importance of Data Ethics, along with best practices for startups to mitigate data risks. Bose is the founder of Ethical AI Governance Group (EAIGG) and a managing partner at BGV.
The Case for Data Ethics
Multiple data points indicate that AI is reaching a new level of maturity.
If AI is the engine for value creation across multiple industry sectors, then data is its fuel.
Along with the rise data comes a concurrent debate over its ethics.
While companies have begun to address the operational aspects of data management (including how to build and maintain a data lake or how to integrate data science and other expertise into existing teams), far fewer companies have systematically considered and begun to address the ethical aspects of data management.
For instance, these issues might include scenarios in which algorithms are trained with biased data sets or data sets are breached, sold without consent, or otherwise mishandled. In these cases companies can incur significant reputational and financial costs. Some of the common “data ethics” traps that enterprise can fall into include:
- Thinking in “silos”: The mistaken belief that the legal and compliance departments have carefully considered Data Ethics or that data scientists have all the answers.
- Examining only the data and not the sources: Ethical lapses can occur when executives look only at the fidelity and utility of discrete data sets and don’t consider the entire data pipeline. That might include the origin of the data, whether vendors received informed consent for use by third parties, and whether the market data contains any material nonpublic information.
In 2021, McKinsey produced a global survey of 1,000 respondents on “The State of AI.” Key findings in that report included:
- Just 27 percent of respondents reported that their data professionals actively check for skewed or biased data during data ingestion.
- Further, only 17 percent said that their companies have established a dedicated data governance committee that includes risk and legal professionals.
- In that same survey, only 30 percent of respondents said their companies recognized equity and fairness as relevant AI risks.
The reality is that Data Ethics poses real risks for enterprises. We just need to look at a few recent examples to understand the dangers:
- YouTube discovered troves of videos with profanity and violent themes in its YouTube Kids service, the video-sharing sites’ kid-friendly platform. A video called “cocaine pancakes” reached nearly one million views without being flagged as inappropriate content.
- An investigation by The Markup, a technology watchdog group, found that lenders were more likely to deny home loans to people of color than to white people with similar financial circumstances. Specifically, 80% of Black applicants, 40% of Latino applicants, and 70% of Native American applicants are likely to be denied loans.
Implementing Data Ethics
A holistic approach is essential for implementing Data Ethics. Some key best practices are highlighted below:
The Data Ethics Canvas developed by the Open Data Institute is another best practice, a simple but powerful tool to operationalize AI ethics.
Startup Innovation
Responding to the need, entrepreneurs are establishing startups to make data-centric AI a reality. Their offerings address some of the challenges around improving AI governance and Data Privacy. The chart below displays some of their services.
Conclusions
With the growing adoption of AI, Data Ethics is becoming an increasingly important risk area for Enterprises to address. The stakes for companies could not be higher. Organizations that fail to honestly address data ethics risk losing their customers’ trust, eroding their reputation, taking on legal liabilities and ultimately destroying shareholder value.
We must recognize that comprehensive Data Ethics cannot be put into practice overnight. Building teams, establishing best practices, and reforming organizational culture all take time and resources. We must also recognize that regulation alone is not the panacea:
- The European Union’s General Data Protection Regulation (GDPR) went into effect only in May 2018. The California Consumer Privacy Act has been in effect only since January 2020. And federal privacy law is only now pending in the US Congress.
- Laws can define ethical boundaries for executives, but a comprehensive, internalData Ethics framework can guide executives on whether they should, say, pursue a certain commercial strategy and, if so, how they should go about it.
At the end of the day, honoring your organization’s Data Ethics principles may mean walking away from potential partnerships and other opportunities to generate short-term revenues. But in the longer term, good Data Ethics principles translate into increased trust with customers, an augmented business reputation, and a stronger fiscal foundation.
