Consider these real-life examples:
- Facebook’s M virtual assistant was designed to offer helpful replies and emoticons to liven up messages. However, in a discussion on women’s health issues, it inappropriately suggested a vomit emoji, and it chimed into a discussion about a horror movie scene (don’t ask) by suggesting that the friends who were messaging make dinner plans. M has since been pulled for a re-tooling.
- Using cameras placed around a street fair and a real-time facial recognition algorithm, the Metropolitan Police Service of London searched a crowd of two million for more than 500 people wanted for arrest or barred from attending. By the end of the day, 95 of the 96 matches were false-positives, and the 96th person identified had already been apprehended and released at an earlier time.
- Eager to demonstrate neutrality in court sentencing, a court system incorporated an AI-driven recommendation engine that predicts recidivism. The process is opaque, as it only reveals the recommendation and not the methodology, which the state’s Supreme Court contends does not violate the right to due diligence and discovery (Wisconsin v. Loomis). Later, researchers discovered that the algorithm was far more likely to incorrectly judge black defendants as being at a higher risk of recidivism than white defendants, while white defendants were more likely than black defendants to be incorrectly flagged as low risk.
What do these cases have in common? First, it’s solid proof that AI has moved into the real world, as the theme of the upcoming AI Summit in San Francisco suggests. Second, these examples are early warning signs of how AI and analytics initiatives are headed for structural failure if they are not conceived of, crafted, trained and continuously monitored by humans, for humans, and with human interests in mind.
AI truly is doing remarkable things on a small scale to improve safety, human health and business effectiveness. But without the re–humanization of AI, the point solutions taking root across industries may never reach their full potential. As the examples above demonstrate, AI’s ascent in the real world could be limited by social miscues, embarrassing inaccuracies, misguided legal opinion and damaging racial bias. A real human would more likely be able to analyze, understand and fathom the context, emotion, intent and outcomes of AI-driven algorithms, and avoid these false starts. Therefore, re-humanizing AI must become a business imperative.
Getting the Most Out of AI
So how do practitioners and business champions avoid AI failure and take advantage of AI’s potential? A longer answer is discussed in this article, but let me provide some hints – a few of which will be shared during my AI Summit keynote on September 19. Our success helping clients across industries and continents extract greater value from AI reveals key words to the wise.
- The “I” in AI refers to human intelligence, not rules-based automation intelligence. We believe that anthropology, digital ethnography and behavioral science are a requirement for any meaningful AI program. Understanding what people want – really want – is key to inserting AI’s intelligence in a way that seems natural and helpful.
- The training data that feeds AI originates from real-world experiences accrued over the years, which can often be biased. We overcome this by applying techniques that allow AI to scale and grow with precision and transparency into conclusions and methods. This transparency, when applied diligently to enhance personalization and trust, reduces bias in AI and encourages consumer and employee acceptance.
- The fundamental question, “Do you know if you have the information you need to move your business where you want?” has far too many business leaders take an uncomfortable pause as they think of an answer – and even those answers likely come from gut feel and not from a position of solid data analysis. Companies that know why things happen, and control the underlying processes that drives those outcomes, will leap ahead of those that follow purely quantitative paths to understanding. We call this causality, and it will change the way we think about business analytics and insights.
AI is helping us, not dooming us. AI is indeed working beyond the hype, and is improving the quality of consumer and employee experiences while also lifting the profits and quality of the company’s offerings and services.
No matter where your organization is in its AI journey – whether you’re struggling to start or ready to take the next big leap – it’s essential to re-humanize your AI solutions in order to extract greater value for your business, your customers, and the societies you serve.