How psychology is shaping better machine learning
- 11 September, 2017 07:13
More psychologists are now coming the tech space because they're trying to teach machines to become more social and sociable, according to BT’s head of customer insight and futures, Dr Nicola Millard.
“I’m not a technologist, I’m a psychologist – and it makes a lot of sense having me on-board because innovation in itself won’t work unless people adopt it," she told CMO. "A psychologist in the team prevents us from getting carried away exclusively by the technology which often tech companies do."
Leading the third largest innovation hum in the UK, Millard is responsible for tapping into the research and innovations that BT does for its global services clients.
“I used to have a silly job title as a futurologist, and I hated that job title because everyone assumes you have a crystal ball,” she said. “But I’m in global services so my clients are typically retailers, airlines and banks – big global corporates – and we often bring them around to our showcases of the retail store of the future, or the bank of the future, so they can have a play with our proof of concepts."
While Astral Park in the UK is BT's main research hub, the company also has other centres dotted across the world including Abu Dhabi, Singapore University in Beijing and MIT in the US. Millard said the company also has tech scouts dotted around the world bringing in new startups.
In her role, Millard looks the changing interactions between consumer and customer, and how customers’ demands and expectations on companies are rapidly evolving, particularly if the company is providing a service. The other element is the concept of the ‘digital employee’ and how the changing nature of employees relates to the future of work.
“From my experience, I see a lot of tech that allows us to collaborate, but there’s also too much tech, which opens up the risk of fragmentation, especially when we all have such diverse workforces,” she said. “Successful collaboration using technology rests on who is actually on it. Look at social media platforms for example.
"A lot of what I look at in that area is creating common ground, which I define as it has to be absolutely accessible to everyone and it has to be appropriate to the task.”
On a technical level, Millard sees common ground as looking at disparate technologies and getting them talking to each other in the cloud in an integrated way.
“But what I’m also interested in is how to get leaders to choose which common ground works better for them to communicate and collaborate – and that could actually be face to face, because that can be incredibly valuable,” she said. “At BT, we’re looking at how can the digital space, using tools like video, audio, chats, AR, VR, create as good an experience as that face-to-face interaction.
“We also look at who is using which technologies, what the trends are and then how can we then improve the technology and enhance the end experience.”
AI, machine learning and automating the customer experience
As a psychologist, Millard has undertaken a lot of academic work to understand how to make machines more natural to interact with and how to use technology to create better customer experiences – whether it is in the physical or digital world. As a result, she’s considered in great detail how machines and artificial intelligence (AI) can understand and communicate in natural language as part of an effort to create more seamless bot experiences for customers.
“The way language has evolved isn’t really about rules or process, which is what machines like,” she said. “Machines do simple stuff well, but complicated stuff like understanding regional accents, complex emotional complaints or sarcasm, not too well.
“This is because a lot of things we do naturally, they can’t do naturally. Things like empathy, caring, negotiation, innovation, creativity – and this can present some issues in customer service where we’ve been looking at chat bots.”
For instance, Millard recalled a sarcastic customer complaint against a British train company that said ‘thank you for my free sauna this morning’, which the company’s automated bot interpreted it as a good thing.
“We know it’s a bad thing, but why would a machine know that?” she asked. “It’s the subtle stuff that’s difficult for machines at the moment."
Where AI does work in customer service, from Millard’s extensive experience, is where it has been given the right data.
“There’s a lot of hype around AI and a lot of our early experiments in this field failed largely because of the cost, but what we learned is you need data to make these things work,” she explained. “Some AI tools are working really well for companies at the moment, but they need to be deployed where there’s lot of data, because AI cannot magically create data - it’s only as good as the data you give it.”
“So before you implement it, think about what data you have, what form is it in, and whether it will actually work.”
A simple way to make bots work in your favour is to simply turn your FAQ section of your website into an interactive question and answer bot conversation your customers can engage with to quickly find a solution, Millward suggested.
“You need to think about whether leveraging a bot actually adds value – it might not work on complex complains necessarily customers,” she said. “But if you can translate your FAQs into an interactive chat and the bot answers the questions your customers ask – then it could work as it gets the answer quickly to your customer.”
AI is also currently working well in a customer service ‘triage’ environment, Millward said. While it might not offer all the answers to customer queries, it can direct the customer down the right channel, whether it is to a bot or a huma.
“This can work quite well because it combines man and machine,” she said. “As long as you don’t pretend it’s not a bot, you need it to clearly say ‘I’m a bot, but I’m handing over now to my human colleague’ – you need think about all these subtle elements to enhance your customer experience.”
“But all these things take time to train. I’m working on three bot projects on the moment. It’s complicated, machines need to learn, process the data, and it takes time. And it might not work.”
Digital butler versus online stalker
Millard is excited about what the future holds for machine learning, but she warned companies to steer away from the ‘creepy stalker factor’.
“I like the digital butler concept, that my technology knows me well and does things to help free my time and brain up to do other things,” she said. “What I don’t like is when my technology becomes like a stalker running behind me constantly tapping me on the shoulder to show me things I don’t necessarily want.
“There’s even camera technology now that can look at your micro-expressions on your face and tell you how you feel. But will it all get too creepy? Are the likes of Google home and Alexa finding out too much about me? Is the technology telling me too much that I don’t want to know? Well that’s where psychologists can come in to rally help shape the tech space to make it more sociable and less intrusive.”