AI is prone to errors but will humans forgive the machines?
AI driven decision-making will form the guts of an enterprise or government.
Cricket World Cup is always huge for cricket playing nations. During the 2019 ICC Cricket World Cup, England and New Zealand clashed in the finals after spectacular victories against highly accomplished teams. In a hotly contested game, England won by a whisker, after a controversial decision by the umpire. In the final over of the game, crucial six runs were awarded to England when a ball thrown by a fielder, ricocheted off the batsman's bat and sped to the boundary. Long story short, this stroke of luck or rather the umpire's misjudgment nudged England towards its maiden world cup win.
The game of cricket has amassed cool technologies that have reduced the pressure on umpires. Earlier they had to take split second decisions. In hindsight some of them looked a bit iffy. Along came technology that added transparency and consistency in their decision-making process.
During the world cup tournament, the LED stumps and bales when hit would automatically light up, adding more transparency in run-out and stumping decisions. Leg Before Wicket decisions have always been mired by controversies. The projected trajectory of the ball is so subjective that TV commentators having a near identical line of sight could have opposing opinions. Technologies like ball tracking system and Hot Spot technology have eased the final decision making. The umpire still gets to raise his index finger, but often with a little help from technology. However, there can still be instances of unintended human error in complicated situations of the game. This is what happened during the ICC World Cup.
Human decision-making is based on our fuzzy judgement. Cristiano Ronaldo's soccer skill was put to test in an experiment wherein he was asked to score a goal in pitch darkness. Each time the ball was kicked towards him, he scored a perfect goal. But ask him how he accomplished it by just listening to the sound of the ball, he won't have a good answer. It is hard to precisely know why a soccer player gets it right or wrong. It is a mystery to all of us.
What goes on in the prefrontal cortex of our brain has been extensively studied by neuroscientists. Decision-making involves maximising gains at minimal decision costs. In other words, speed-accuracy tradeoffs. A cricket umpire knows this all too well, so do stock exchange traders.
Many motivations and cognitive biases also lead us to errors in judgment. Besides, our preloaded past experience, our mood matters. The weather is not good, or you are feeling a bit under the weathe - all this could alter your judgement of the same situation. Your boss' reaction to your leave application often depends on his or her mood. Even AI is learning to understand your tone of voice because our mood can go from happy-to-frustrated within minutes. The variability in human judgement is called noise in statistics. Psychologists have confirmed that even professionals contradict their own past judgements.
But if Artificial Intelligence (AI) is presented the same situation, and assuming that it has the same bias as humans, because we programed it (a discussion for another day), it will give the same consistent decision each time. This is because algorithms are noise free. Amazon's Alexa will greet you in the same consistent manner every morning.
AI driven decision-making will form the guts of an enterprise or government. Technologies used by cricket umpires is a great analogy of how humans will use AI in their decision-making process. Umpires still call the shots, but where the human eye is misled and unsure, technology is just what we need - a perfect example of augmentation of human skills.
AI can also play a good third umpire where humans have a difference of opinion on the field. Sometimes, when players request for a review of the umpire's original decision, the ball tracking system combined with replays in slow motion tell us that he was spot on. Everyone goes back to playing the game without any rancor. AI robot judges will start to resolve small claims cases. With over 4 million cases pending in its 25 high courts, India urgently needs an AI robot judge to clear this massive backlog. US courts use algorithms in sentencing criminals.
Government can also make use of AI to make rational decisions on subsidies, scholarships, funding and social benefits. AI can consider more complex all-round nuanced factors. But we obviously need more transparency in how AI comes to a decision.
With this comes a huge responsibility for AI. Humans are extremely understanding about human error. The philosophical phrase, 'we are human, after all' is how we forgive and often forget. We are allowed to make mistakes. When the dust settled after the ICC world cup, the losing side was gracious enough not to rake up the umpiring error. The cricket governing body ICC tweaked the rule book, while the world moved on to other pressing matters. But we will not be so forgiving of AI making mistakes, because we will expect the machines to get it right. When it comes to AI-driven decision-making, the risks and rewards are equally high.
Shalini Verma is CEO of PIVOT technologies