EaaS — Can Empathy Be a Service
There was an old joke said by a customer I worked with. “One day you will have EaaS, where E stands for everything. Platform, hardware, license, developers, oil drilling machines- everything can be rented as a service”. We had chuckled back then even though things like MBaaS, IaaS etc had already taken shape, which lent some credibility to the customer’s rant.
That was then. Recently I stumbled upon the great book on human and machine collaboration, “Human + Machine: Reimagining Work in the Age of AI” by H. James R. Wilson and Paul Daugherty (https://www.accenture.com/in-en/insights/technology/human-plus-machine)
And that is where I encountered the topic I am speaking about today- having Empathy as a service. It struck me as wondrous and yet, something sensible and meaningful for our times. Let me explain.
My friend had lost her credit card courtesy stolen wallet. She frantically called up the bank to get her card blocked. The AI-enabled IVRS greeted her. The IVRS was designed smartly enough to offer the theft/block option as the very first in its list. That was good.
The friend said “theft, yes stolen. Please block my card” with unconcealed anxiety. The IVRS moved to a different menu. While that was okay, there was no… emotional acknowledgement of the problem. Just the standard robotic, “Oh we are sorry” said like an afterthought. There was no urgency in the options presented in the subsequent menus, while my friend tried to frantically speak out the correct option as soon as she heard it. Her problem stood solved at the end of a 15-minute phone call, but her state of mind was only marginally better off.
In another time, I found an elderly couple trying to book flights off this renowned travel portal. They had been convinced by their son that speaking to the AI-enabled call center was the best option since they were not comfortable using apps and typing on small screens. I offered to help but they wanted to give it a try.
I watched them first try the chat-bot on the web to try to get the system to call them. While the chat-bot seemed more powerful than a FAQ navigator, it did not bother to know about who it was speaking with. When the call finally connected, the obviously AI-sounding call center system started off by telling them whether they wanted flights or cabs or hotels. They were enticed by what they heard but I heard the gentleman whisper, “What do we say if we want to book all three?” “Just, just say flights for now”, said his wife. They were also interested in extra leg space and luggage allowance, but somehow the call-centre got to that a lot later. It was also tricky for the couple to remember the various flight options being presented. I almost thought that a human there might have said, “You can take a minute to write this down or I am marking this for you as a potential selection”.
Today, AI proliferates a lot of corners of our lives whether trying to book a service, get some counselling, buy groceries or simply to know what time it is. AI is also getting to help humans be better at their jobs by amplifying their capabilities, whether it is helping us better manage schedules, perform tricky surgeries or recommend the best product to our customers. Intelligence in systems is now in ample supply. Not as ample as some visionaries would say, but still replete with opportunities to improve our life and our abilities.
I believe that the next step in AI, whether in chatbots or cobots is the presence of empathy. A recent demonstration from Google’s voice assistant showed how it captured nuances in human voices and responded with the right words and verbal gestures.
There are systems that are being trained using deep-learning to sport better empathy. But what if it could be a service? Building empathy in systems takes tremendous amounts of data and training. Not everyone investing in AI can also invest in making the system empathetic. What if, we could offer the empathy element of AI in the form of a pluggable library or a web-service?
That is obviously tricky. A Machine Learning engineer I know (though I know Machine Learning is only one of many types of AI) said that deep down a ML algorithm is just a big nest of if-then statements. Then how would you insert new if-then statements in between the existing ones and still make sense? Maybe that’s not feasible. Or maybe the entire suite of AI services can be made into a service. Or the idea may be to think of entirely different architectures to implement AI. In my head, the promise of EaaS is far too great to limit myself on account of feasibility constraints (as valid as they may be).
One idea could be to have a filter which accepts inputs as the standard AI-response and information about who the receiver is and sends back a revised response which is empathetic.
As an example, the bank AI system sends the following as the inputs,
Standard response: “Oh, we are very sorry. Can we help you?”
Customer information: “Female, aged 26, stays in India” etc.
And the response from EaaS could be,
“Well, that just sucks. It is very stressful to lose your card and risk losing your hard earned money. But we will try our best to help you. Can you help us with some information?”
The response need not just be text. The EaaS service could also return parameters like voice pitch, pause durations between phrases, articulations and fillers which can then be understood by the AI-system to modify its responses.
Suddenly, it’s not just Google voice assistant that is empathetic but rather anyone who wants to use Google’s Empathy AI service to bolster their own AI-offering. EaaS would elevate your standard run-of-the-mill AI responses and robotic gestures to be more human and closer to the customer’s world. Voice is just one use case. The EaaS system can transform your texts, your storeroom designs, your magazine covers and your cold-calling scripts. You can feed it back as well to train and re-train the EaaS and your base system.
With EaaS, AI moves closer to our world and our problems. It moves beyond mimicking us to actually caring about us. And that can transform the way we perceive AI and the value it provides to us.
Who do you think about EaaS and its potential? Let me know in the comments!