AI, Big data, Automation,
And That Is About To Get Scarier
Artificial intelligence already decides who you are. Your reputation online influences whether people date you, hire you, buy from you, rent to you, loan you money, admit you to their school and much more. Many of these decisions involve subjective and highly personal calculations. It will impact lives directly and in deeply personal ways.
New AI will determine your reputation and other AI algorithms will learn about you from that information and make decisions about you. This means that a lot about our lives will depend upon competing versions of interconnected AI, each of which reflects the professional interests and biases of the AI owners and developers. These interests are focused on maximizing the AI owners benefits and that can be at odds with your pursuit of happiness.
Google is one of the top AI researchers and has used it to improve search results for a few years. Their RankBrain algorithm evaluates new content about you. This allows Google to deliver more relevant search results than traditional techniques like keyword analysis and website link scoring. Two thirds of people trust online searches for research more than any other source for things like purchasing decisions, job candidates and dating.
Google optimizes search results to maximize users’ clicks. This is because Google interprets that as reflecting what searchers find relevant. But that doesn’t necessary align with what people want others to see about themselves. A doctor may have delivered a thousand babies and spend every weekend scooping soup at a homeless shelter. But, if a news article about her DUI arrest from 1999 gets the most clicks then that is what Google will display first for her name.
Google AutoComplete, which makes suggestions that pop up as you type in the search bar, flavors opinions even when it doesn’t reflect actual content. Search activity determines those suggestions. Soon predictive algorithms will determine what is relevant about you and make suggestions on its own. That shift is important. It moves us from recommending things based on human activity to AI determining what will be or what should be.
Dating sites will be heavy users of AI. Tinder chairman Sean Rad announced their plans to use artificial intelligence to help reduce the noise and effort of searching profiles. This means that Tinder will influence even more who its members see and therefore date. The next step is to predict whether you are looking for casual hookups or serious relationships at a given time of day and week or stage of your life. If you think your mother telling you it’s time to settle down is intrusive, imagine Tinder telling you that it thinks you are ready to settle down so it found a match for you.
Recruiting is a valuable but labor-intensive and error-prone field. Jobs are vacant an average of one to two months and a bad hire can cost a company more than three months’ salary in losses. Most recruiters say finding the best candidates in a big database is the largest part of their job.
Applicant tracking systems go beyond keywords to predict who will make the best candidate. Karen.ai is a cognitive recruiting chatbot that develops personality insights from interactions with applicants. Recruiting companies get paid when a candidate is hired. So, much like dating apps, they are focused on maximizing successful matches by predicting who the hiring managers will hire. Corporate recruiting platforms share that but also have incentive to predict job performance. In either case, these platforms are not there to find you a job or maximize your opportunities. So, if your profile doesn’t appeal to AI you are just not going to find work.
Online review sites are starting to use AI to improve the quality of customer reviews, which are prone to mischief and errors. Businesses with predominantly one or two-star reviews lose as much as 90% of sales prospects. As more people join the gig economy, reviews and ranking features of freelancer platforms and peer-review sites will determine a lot of hiring decisions. AI will interpret reviewer intent and motivation to weed out inaccurate or malicious reviews and improve reliability..
The healthcare industry is a major consumer of AI and analytics to improve care and to reduce complexity and cost. Companies are already using AI to predict medical conditions based on patient medical records, their behaviors and symptoms. Some are providing direct feedback to patients in real time to influence behaviors. Here AI arguably may resemble “big mother” more than “big brother.” Car insurance companies experimented with tracking devices in cars to reward safe driving. A next step in healthcare would be analyzing Facebook or Snapchat pictures and posts to look for weight gain, risky pastimes or signs of depression.
In the justice system, prosecutors and defendants are not looking for justice, per se. Prosecutors are looking for successful prosecutions and defendants are looking for exoneration. Prosecutor AI will help them determine strategy for each case. So, what if the AI says that a rich, white suspect is more likely to reject a plea bargain, put up a rigorous defense and be exonerated than a younger, poorer minority suspect? AI will also inform parole decisions which focus on reducing recidivism, so biases matter a lot.
Of all the discussion of artificial intelligence and automation, how it determines things about us gets surprisingly little coverage. This may be a sign that we are still not ready to face our deepest fears of AI. Job automation may feel inevitable because of technology but it isn’t personal. If AI replaces your driver or even your doctor, we trust it is be because the data proves it is better for us.
A passport photo screening system rejected a picture of an applicant of east Asian descent because it interpreted his eyes as closed. Google Image search tagged at least one photo of an African American in a highly offensive way. It is wrong to write these off as bugs in early versions of technology. They reflect an important truth that, like children, AI pick up the biases of their parents.
The argument that AI will eliminate biases in making decision about people misses a critical point. The owners and developers of AI will eliminate biases if there is a benefit to make that a priority. AI supercharges decisions about people and that means some biases could be magnified instead of eliminated.
Reputation is different than other optimizations. It is both quantitative and qualitative and your reputation and the decisions it influences are deeply personal. There is great opportunity for misalignment between your interests and those designed into the AI algorithm. Special care should be put into AI that influences our reputations or makes subjective decisions about us.
Michael Zammuto is the CEO of Completed.com and Cloud Commerce and a strategic adviser to several startups. Mike's background is in SaaS services, B2C sites and B2B firms and has worked extensively in online reputation, digital marketing and branding.