Follow Datanami:
January 27, 2020

The Current State of AI Bias

Fed by massive amounts of data – data that increases in volume every day – AI solutions have the potential to yield unprecedented insights that can be used to steer business, government, and societal progress.

DataRobot surveyed more than 350 U.S. and U.K.-based CIOs, CTOs, VPs, and IT managers involved in AI and machine learning purchasing decisions. Respondents offer a look into how AI is being used by businesses today, current perceptions of AI bias, and what is being done – or should be done – to enhance AI bias prevention efforts in the future.

Key Report Findings

  • 42% of organizations are “very” to “extremely” concerned about AI bias. Most respondents cite “compromised brand reputation” and “loss of customer trust” as the great cause for concern.
  • 83% of respondents have established AI guidelines and are taking steps to avoid bias
  • 85% of respondents also believe AI regulation would be helpful for defining what constitutes AI bias and how it should be prevented.

Primary Concerns about AI System Use

Across the U.S. and U.K., the greatest concern organizations have when it comes to using AI systems is software bugs: 53% of U.S. respondents and 58% of U.K. respondents cite this as a top concern. Other AI system concerns differ by geography. While U.S. organizations worry more about lack of accuracy, U.K. companies are more concerned about unethical uses of AI.

This discrepancy we see between U.S. and U.K. respondents when it comes to their concerns about AI could relate to the regulatory and cultural state of affairs in each geography. While the U.K. has elevated bias and AI trust as federal issues – evidenced by the General Data Protection Regulation (GDPR) – the U.S., for the most part, only operates against recommended data guidelines.

Current Perceptions of AI Bias

While U.K. companies are most concerned about technical AI bias, U.S. companies say that emergent AI bias is most concerning. Despite different takes on the types of AI bias that are most concerning, U.S. and U.K. agree that the most concerning ramifications of AI bias are compromised brand reputation and loss of customer trust. This concern is warranted – and one that is recognized behind even the largest AI players.

How Organizations Are Mitigating AI Bias Today

To get ahead of AI bias issues, organizations are devising strategies to ensure that the AI systems deliver accurate conclusions, enable smart decisions, and can be trusted to foster progress. To maintain customer trust and brand loyalty – and ultimately compete in today’s customer-centric business landscape – companies need to make AI trust initiatives a priority. Most AI bias prevention initiatives today are overseen by a member of the C-suite – most commonly the CIO.

Conclusion

AI bias is a real concern for today’s organizations, and for good reason. Organizations can enlist support from experts – vendors, consultants – to get smart about how AI bias can occur and identify human-friendly AI solutions that offer visibility into how models are created and maintained over time. While AI regulation may be a longer-term reality, soliciting expertise will help organizations feel confident that they’re leveraging solutions that are ethical and trustworthy. Taking a proactive approach will help organizations mitigate risk, avoid disruptions, and thrive once these regulations are inevitably enacted.

Want to learn more? Download our ebook, The State of AI Bias in 2019, to find out about current perceptions of AI bias and how organizations are using AI today.

 

Datanami