White House Big Data Recommendations a Start, but More Data Sought

U.S. Chief Technology Officer Megan Smith, U.S. Chief Data Scientist DJ Patil and Domestic Policy Council Director Cecilia Muñoz said in a blog post announcing the report's release that big data can be used either to advance civil rights or to undermine them, and the government needs to better embrace this technology to provide greater opportunity. The report used case studies on credit lending, employment, higher education, and criminal justice.

"This report is an important acknowledgment of the critical need to pay careful attention to ensure that big data protects and advances, rather than undermines, civil rights," said Wade Henderson, president of the Leadership Conference on Civil and Human Rights, a coalition of more than 200 national organizations, in a statement. "We applaud this emphasis on civil rights in the context of big data, and urge the administration and federal agencies at all levels to do even more to ensure that technological progress brings greater safety and economic opportunity to everyone, and especially the most vulnerable populations.”

Joshua New, policy analyst with the Center for Data Innovation, said the report shifts "the conversation from 'let's be wary,' to 'big data is the solution to these problems ... not the problem itself, so let's figure out how we get it right.'" While the White House's 2014 big data report acknowledged the opportunities for economic and social benefits, it disproportionately focused on potential harms rather than maximize the technology's promising benefits, he said. In analyzing that report, his group identified 37 potential harms listed, but only two cited actual evidence that consumers were harmed, he said. The FTC report on big data released in January (see 1601060042) was better in touting benefits, but it also focused heavily on hypothetical harms, "not to say there aren't potential harms," he added.

The change in mindset with the new report might be due to the administration's hiring of a chief data scientist since 2014, representing the willingness of lawmakers to better understand how such systems work before deciding how to regulate them, New said. The White House also launched promising big data initiatives in education, housing, policing and other sectors to combat bias and discrimination, he added.

The recent report listed several broad recommendations that called for more research into reducing "algorithmic discrimination" against low-income and other communities as well as greater study in the auditing and testing of big data systems. Not only should these systems be more accountable and fairer to consumers, but companies and organizations that use the technology to make decisions about consumers also should be held accountable, the report said. It added that more collaboration is needed among computer and social scientists, among others, and that new standards are needed within the private sector and in the government.

Challenges in each sector will be different and New agreed that more research is needed. "The world is already permeated with human bias whether it's subjective or overt and now that we can collect all this data and put it to good use it's the first opportunity ... that we can kind of push that back," he said. "There will be hiccups along the way, but the only way we can do that is with a lot of data and really, really robust tools to do so." He said the White House's recent announcement it will hold workshops on artificial intelligence and machine learning (see 1605030058) could be another step in advancing the research and thinking in this field.

The report provided case studies on the opportunities and shortcomings in using credit lending, employment, higher education and criminal justice. For instance, the report said companies could cast a wider net to collect more information about consumers with "thin files," meaning they have too little credit history for an algorithm to generate a credit score. Some companies could look at data from phone bills, public records, previous addresses, educational background and tax records that could be useful in assessing credit risk, the report said. It said other companies could tap "less conventional sources" such as cellphone location data, social media platforms, buying preferences from online shopping histories "and even the speeds at which applicants scroll through personal finance websites."

While companies could use a tool to expand credit access to an underserved population, that same tool could be used to "reinforce disparities" and consumers may not be able to dispute such outcomes, the report said. "When such decisions are made within computationally-driven 'black box' systems, traditional notions of transparency may fail to fully capture and disclose the information consumers need to understand the basis of such decisions and the role that various data played in determining their credit eligibility."

World Privacy Forum Executive Director Pam Dixon said her group issued a 2014 report about predictive consumer scoring and algorithmic discrimination, and over the past two years has been digging into usage of big data across various sectors and found unfairness, and that's what the government needs to do as well. “We need to look for very specific instances of problematic uses and really, really go after how those can be fixed,“ she said, especially around academic admissions processes, use of marketing, and alternative data for any kind of consumer scoring in the financial and healthcare sectors.

It's unlikely Congress would act on the concerns that the report raises especially during a presidential election year, but Dixon said it's good to have such ideas documented. "What really speaks to me about the White House recommendations is the ability for people to correct the records and have access to the records," she said. "If you look at marketing data about people there is a lot of inaccuracy in the data and the [industry] argument to date has been, it’s just marketing and it’s not being used for eligibility. And what I have to say is we are in the post-eligibility era. And marketing data is being used for quasi-eligibility decisions and meaningful marketplace decisions that really impact people’s lives."