The topic of bias gains currency each day. As AI becomes more pervasive, we’re seeing more examples of how AI delivers value but also can spread harm through bias inherent in the data sets that businesses use to train AI applications. And these fears are well-founded. It’s easy to find them:
- A University of California-Berkeley study revealed that lenders charge higher rates to Black and Hispanic borrowers. According to the study, algorithmic strategic pricing uses machine learning to find shoppers who might do less comparison shopping and accept higher-priced offerings. This algorithm is biased against Blacks and Hispanics.
- Including the word “transgender” in video titles has resulted in YouTubers receiving lower ad revenue on their videos. Commented Meg Green, a user experience researcher for Rocket Homes, “Being gay or being Black or being a trans woman does not mean these things are negative and that you don’t want to read this information. Anything about being bisexual and gay is pornographic and not acceptable for children, according to some biased data found with AI.”
- A recent study showed that broadly targeted ads on Facebook for supermarket cashier positions were shown to an audience of 85 percent women, while jobs with taxi companies went to an audience that was approximately 75 percent Black. Miranda Bogen, a senior policy analyst at Upturn, noted, “This is a quintessential case of an algorithm reproducing bias from the real world, without human intervention.
Blaming data scientists for bias is not the right solution. Finding a way to combat bias at a systemic level is.
But what is the answer to rooting out bias? Unfortunately, I see too many answers blaming people, especially data scientists. According to a popular and misguided rationale, the data scientists who create the data sets that power AI-based applications humans suffer from biases, consciously or not, and they bring those biases into the creation of data sets. As a result, AI algorithms unwittingly reinforce bias.
But focusing on the human beings who work with AI data sets overlooks a bigger problem: bias in data sets results in systemic bias at a societal level. Individuals simply reflect that bias. They don’t create it. Bias permeates every aspect of our lives, from our laws to our ways of working – from how we get healthcare to how we communicate. Putting the focus of bias on human beings places an unreasonable burden on data scientists.
Instead, we should recognize that a systematic problem requires a systemic solution. This idea is gaining currency. For instance, recently, Jeff Shuren, director of the FDA’s Center for Devices and Radiological Health, said the healthcare industry needs methodologies for the identification and improvement of algorithms prone to mirroring systemic biases in healthcare.
Mindful AI is one such solution.
Mindful AI is Ethical AI
Mindful AI is an approach to develop AI solutions that are more valuable because they are relevant and useful to the people they serve. To be more relevant and useful, those solutions need to be as free of bias as possible. Mindful AI does not eliminate that bias, but it mitigates against bias harming a solution built on AI. Mindful AI has three critical components:
1. Mindful AI is Human-Centered
From the inception of the AI project, in the planning stages, the needs of people must be at the center of every decision. And that means all people – not just a subset. That’s why developers need to rely on a diverse team of globally-based people to train AI applications to be inclusive and bias-free.
Crowdsourcing the data sets from a global, diverse team ensures biases are identified and filtered out early. Those of varying ethnicities, age groups, genders, education levels, socio-economic backgrounds, and locations can more readily spot data sets that favor one set of values over another, thus weeding out unintended bias.
For example, when applying a mindful AI approach and leveraging the power of a global talent pool, developers can account for linguistic elements such as different dialects and accents in the data sets.
At a recent Fortune Most Powerful Women Summit, speakers underscored the value of putting in place a diverse team of people. For instance, Lisa Edwards, president and COO of Diligent, said that having people from diverse backgrounds in data science governance positions prevents a situation where a group’s bias simply reinforces the technology’s bias. Mindful AI extends that notion to the resources that companies use to perform data collection and curation.
Establishing a human-centered design framework from the beginning is critical. It goes a long way toward ensuring that the data generated, curated, and labeled meets the expectation of the end-users. But it’s also important to keep humans in the loop throughout the entire product development lifecycle.
Humans in the loop can also help machines create a better AI experience for each specific audience. AI data project teams, sourced globally, understand how different cultures and contexts can impact the collection and curation of reliable AI training data. They have the necessary tools they need to flag problems, monitor them, and fix them before an AI-based solution goes live.
Human-in-the-loop AI is a project “safety net” that combines the strengths of people – and their diverse backgrounds with the fast computing power of machines. This human and AI collaboration needs to be established from the beginning of the programs so that biased data doesn’t form a foundation in the project.
2. Mindful AI Is Responsible
Being responsible is to ensure that AI systems of free of biases and that they are grounded in ethics. It is about being mindful of how, why, and where data is created, how it is synthesized by AI systems, and how it is used in making decision that can have ethical implications. One way for a business to do so is to work with under-represented communities to be more inclusive and less biased. In the field of data annotations, new research is highlighting how a multi-annotator multi-task model that treats each annotator’s labels as separate subtask can help mitigate potential issues inherent in typical ground truth methods where annotator disagreements may be due to under-representations and can get ignored in the aggregation of annotations to a single ground truth.
Trustworthiness comes from a business being transparent and explainable in how the AI model is trained, how it works, and why they recommend the outcomes. A business needs expertise with AI localization to make it possible for its clients to make their AI applications more inclusive and personalized, respecting critical nuances in local language and user experiences that can make or break the credibility of an AI solution from one country to the next. For example, a business should design its applications for personalized and localized contexts, including languages, dialects, and accents in voice-based applications. That way, an app brings the same level of voice experience sophistication to every language, from English to under-represented languages.
There Is No Easy Fix
If it sounds like Mindful AI is complicated – well, it is. There are no easy solutions to fixing systemic bias in AI. On the other hand, we know that asking human beings to simply hold each other accountable is not reasonable. At a personal level, yes, we can and should hold each other accountable, but asking people to figure it out themselves will not address the problem because no individual can fight bias in AI. Human beings need help at a systemic level. Mindful AI is one way to provide that help.