AI report shows concern for justice and fairness

Artificial intelligence has come a long way. From fantastical film depictions that are divorced from reality, to cautionary tales of robots taking over the world, artificial intelligence has largely been fiction. But artificial intelligence is no longer relegated to fantasy and conjecture, as technology has moved into the world of automation. Indeed, it is so mainstream that earlier this month the White House issued a report titled, "Preparing for the Future of Artificial Intelligence."

Among other things, the report provides a history of artificial intelligence, including the use of artificial intelligence to support Apple's Siri and the victory of artificial intelligence in a Jeopardy game show. It considers how the increased use of artificial intelligence may affect jobs and the economy, opining that automation will likely affect lower-wage jobs the most, while potentially increasing opportunities for more educated workers. The report further considers how artificial intelligence will be regulated. (For example, whether existing regulations governing the automotive and aircraft industries are broad enough to cover the use of artificial intelligence in self-driving cars and planes.)

The report's findings are distilled into 23 recommendations that seek to foster research and development into artificial intelligence, promote the use of technology based on artificial intelligence, and caution responsible deployment of artificial intelligence.

One particular area of concern is highlighted by the report - the use of big data. Observing that "AI needs good data," the report examined how big data drives artificial intelligence by informing decision-making processes. Humans rely on their knowledge and experience to make decisions. Likewise, technology using artificial intelligence has to rely on the datasets available to it. As the use of artificial intelligence proliferates, the data collected through such mechanisms will in turn be added to the large big data bucket, self-perpetuating any internal biases that may exist in the big data underlying artificial intelligence. Understanding this, the report expressed concern over how artificial intelligence may affect justice and fairness. Indeed, the report cautions that artificial intelligence should promote "justice and fairness, and that AI-based processes [should be] accountable to stakeholders."

This is not the first time the White House has opined on big data. In both 2014 and earlier this year, the White House issued reports focused on the issue. And the Federal Trade Commission recently issued its own report on how big data may inadvertently exacerbate biases and contribute to discrimination.

So what does that mean? For companies developing technologies that rely on artificial intelligence - and, by definition, the big data that drives artificial intelligence - it's important to consider the report and its concerns about justice and fairness. For example, if you provide computer systems and software that can assist judges in making "risk predictions" of certain convicted criminals, consider whether the data being fed into that system incorporate biases, including those based on race, religion and gender.

If you build artificial intelligence into products that allow employers and landlords to screen job applicants and housing applicants, you should take care to avoid inadvertently making decisions that would violate discrimination laws such as the Fair Housing Act. For example, screening through technology relying on big data may disparately affect the disabled, even if the disabled status is not considered directly.

If you're offering consumer loans, consider whether your loan calculation systems incorporate biases based on protected characteristics. In other words, while artificial intelligence will significantly streamline how companies make decisions, human discretion may otherwise be required to prevent running afoul of any laws (and the concurrent public relations disaster).

These concerns over the use of big data will not just affect the users of big data, but it will likely affect big data companies and those who provide analytics of big data. For example, those who rely on artificial intelligence and big data may push their responsibilities down, contractually requiring big data companies to essentially filter their data to prevent the types of biases that can lead to the injustice described by the report. As artificial intelligence becomes more widely deployed, there will also be more pressure on big data companies to be more transparent on where they get their data, how that data is processed, how that data is used, and how that data is shared.

As artificial intelligence continues to expand, companies involved in artificial intelligence - and the big data that drives artificial intelligence - should monitor such regulatory efforts and adapt accordingly. The White House has cautioned that those who use artificial intelligence should be held "accountable," and other regulatory bodies have echoed similar concerns. In other words, while the government wants you to embrace artificial intelligence, you should consider using some human discretion in doing so.