Topics In Demand
Notification
New

No notification found.

Is Synthetic Empathy Holding Your AI Back?
Is Synthetic Empathy Holding Your AI Back?

June 8, 2021

AI

19

0

Human conversation is nuanced. AI is efficient. Getting the right balance of human and AI in support conversations is tricky  — but critical. 

“Are you a bot?” I asked. I had a question about a car I was researching. The bot answered my question but threw in something extra, saying, “It’s a sweet ride.” I found this mildly annoying because I suddenly could not tell whether I was speaking to a human or a bot.

Is Synthetic Empathy Holding Your AI Back?

If it was a human, I might ask different questions. If it was a bot, I really only wanted facts, not synthetic empathy.  AI can deliver answers faster and better than humans in many situations. That is the AI superpower.

Trying to make AI more human by slowing down conversations and adding in empathy when a customer knows none exists is a bad idea that diminishes the experience and annoys the customer. 

Why Synthetic Empathy Annoys Customers

Synthetic empathy is when an interaction designer endows an AI with response patterns designed to mimic human empathy.

We do this as a nod to the human tendency to constantly pad conversations with empathetic touches designed to ensure that we do not hurt someone’s feelings and to show we are paying attention to them.

I hate the fakies — how about you??

By following human conversational protocols, the communication adopts a less direct path to the resolution of a problem or question. This is a grave mistake. Forcing a circuitous path to mimic humans dilutes the superpower of AI and upsets the people it is trying to delight — your customers.  

Human conversation

Human conversation is perhaps the most inefficient communication protocol on the planet. The redundancy and empathy that form the core of normal conversation is deeply ingrained in person-to-person interactions.

We are conditioned to expect communications to happen in a certain way when we attempt to solve problems. We expect that rebooking a flight will require us to get on the phone, explain our situation, and experience an agent saying, “I’m really sorry you missed your flight. I will help you rebook.”  

That said, almost all of us would prefer saying, “Hey, Siri. My flight was canceled — rebook me on the next flight,” and having it happen instantly.

We would not recall nor care about pleasantries or empathy because our problem was solved quickly and efficiently. In this case, Siri and the AI behind it respected our time. That would actually demonstrate respect for our time — and in reality — that conversation would show the most profound demonstration of empathy.

Conversely, most of us feel annoyed when we are talking to a conversational AI support agent and it says, “Hmm, give me a second to find your records,” and inserts an artificial pause. This is a classic example of “synthetic empathy” that needlessly mimics human communications. 

Synthetic empathy can mirror — but fails to deliver truth.

Synthetic empathy can very closely mirror human empathy. But the synthetic version invariably fails to deliver on the intended effect of increasing customer happiness because it feels fake.

An even greater sin is that synthetic empathy too often injects needless friction into interactions and slows down the completion of a task.

For example, a complicated set of tasks such as troubleshooting a streaming video subscription, updating the new settings, and crediting a customer back for missed time might take 15 minutes of human interactions. The task is a couple of seconds task — not 15 minutes for an AI bot.

Speed — Speed — Speed

In this case, the best and most desired form of empathy is the speed of execution. I don’t care whether an AI lets me know that a car I am researching is a nice car. I want to ask the next question, thank you very much. 

In addition, synthetic empathy not only limits the ability of the AI to meet our needs as quickly as possible — but also redirects AI product design down a rabbit hole of human mimicry.

Once we remove the constraints of maintaining a human-like conversation, then AI becomes far more powerful and useful for humans. Product designers can think in terms of speed of resolution in AI terms rather than conversational patterns in human terms. 

The Data Says Synthetic Empathy Is Not Welcome

Synthetic empathy is rampant in AI systems today due to the belief that cognitive systems should echo the way we talk to each other.

In fact, the data doesn’t support this belief. When we surveyed X number of people about their hopes and expectations for customer service, they consistently placed speed, convenience and efficiency as their top desires.

What does your customer REALLY want?

Customers want issues to be resolved quickly and preferred not to wait. And this is a key point.

A majority of these respondents said it is crucial that companies respect their time. Granted, no customer wants services, sales, or support interactions that are rude and unpleasant. But, all things being equal, programmatic synthetic empathy will not move the need for customers — no matter how close to real-life conversations the AI can achieve.

A New Context-Dependent Definition of Empathy for AI 

This begs a bigger question. Is empathy context-dependent?

That data seems to indicate the answer is “yes,” in particular as we consider machines performing cognitive tasks. For this reason, we need to update our definition of empathy to account for the strengths of AI. And we need to avoid embedding synthetic empathy into AI.

Taking this a step further, we need to rethink customer empathy and embrace a new definition of AI empathy that prizes the timely and efficient completion of any given task. 

Customers want speed and convenience.

AI can do things in a tiny fraction of the time that a human brain takes to synthesize and communicate the task. In fact, AI can easily anticipate our needs. A really empathetic AI might note that your flight was canceled, rebook you on the next flight out, and send a text with the new flight info and a note that this is the fastest way home.

There are no pleasantries or pregnant pauses to show how human the AI is. The AI demonstrates a higher level of empathy by saving us time and hassle.

All of the desired attributes that customers want from an interaction with a company are speed, efficiency, complete resolution. These are true attributes, and they describe the improvements that efficient, focused, and task-oriented AI can bring to customer interactions.

Yes, we want AI — Just let AI do actual AI work

Only an AI can note your flight was canceled, simultaneously analyze six different flight options and instantly compute which one is the best option to rebook on to get you home as quickly as possible.

A human cannot process information as quickly and is limited by their senses. So why design AI systems with human limitations rather than AI superpowers? 

With this new definition of empathy, we can start to rethink what customer-facing AI does and how it does it. We can shift from reactive and even predictive to proactive, anticipating multiple wishes and quickly delivering them all.

With the airline example — the AI will not just rebook the flight but also schedule an Uber and text your family the new arrival information.

We can ask AI to optimize for its superpower — speed and personalization.

Programming AI for speed and personalization will make AI contextually more sympathetic by removing the synthetic version.

The gift of time and convenience is the best way to show a human you really care about them.

Sources/References/Author: Puneet Mehta


That the contents of third-party articles/blogs published here on the website, and the interpretation of all information in the article/blogs such as data, maps, numbers, opinions etc. displayed in the article/blogs and views or the opinions expressed within the content are solely of the author's; and do not reflect the opinions and beliefs of NASSCOM or its affiliates in any manner. NASSCOM does not take any liability w.r.t. content in any manner and will not be liable in any manner whatsoever for any kind of liability arising out of any act, error or omission. The contents of third-party article/blogs published, are provided solely as convenience; and the presence of these articles/blogs should not, under any circumstances, be considered as an endorsement of the contents by NASSCOM in any manner; and if you chose to access these articles/blogs , you do so at your own risk.


© Copyright nasscom. All Rights Reserved.