H10

Confessions of an artificial intelligence psychologist

The first confession in this piece is that we never assume anything at Jade. So if you have somehow had your head in the sand for the past 20 years, it’s worth covering off exactly what is artificial intelligence is.
Jonathon Bray

Artificial intelligence is basically a computer system that has learned how to solve a problem, based on large sets of data, in which it makes assumptions based upon common patterns found - although we never make assumptions, it’s a different story with AI).

While we’re on the subject of covering things off, it’s worth defining machine learning, which often gets mixed up with AI. Machine learning is a building block for AI. It gives systems the ability to learn, identify patterns and make decisions based up predefined sets of data. 

 

Psychology and artificial intelligence - an interesting combination.

Cambridge Dictionary defines psychology as “the scientific study of the way the human mind works and how it influences behaviour, or the influence of a particular person's character on their behaviour. Google gives an extra little tweak, which is aligned to Jade's approach - the scientific study of the human mind and its functions, especially those affecting behaviour in a given context.”

AI and psychology are related due to their core functions of defining patterns and predicting outcomes. The lens that we view AI through at Jade work is conversation - a digital human talking to a real human. Another way to think about this is that AI is the perfect human assistant.

It’s important to note that AI is fundamentally a machine. It can analyse text sentences and identify emotional sentiment, all the while remembering that context in a conversation can blur those lines. Humans are complex and every scenario can be different. So we’re always building in sensors that can identify the appropriate moment to hand the conversation over to a human.

In short, AI and psychology mix together well in a couple of ways. One is the similarity between behaviour and pattern recognition. The other in a user’s expectations and how they interact with machines. 


How people are interacting with digital employees. 

There’s still a massive hangover from the early chatbot days. In fact, some companies are still making, delivering, and using exactly the same ‘chatbots’ that everyone loves to hate. This attitude exists because the experience was never really considered in the beginning. It was a technology solution. That’s why we take a psychological approach, being mindful of the behavioural context and the experience people expect. People deserve better. 

From the studies and workshops we’ve carried out with intelligent digital employees, we've noticed some interesting behaviour from some participants. People understood it was a bot and therefore their expectations were at times quite low, or rather sceptical. We asked users something along the lines of “Tell me about yourself? I'm interested in things like your age, if you smoke, own a home and have kids” - these people were quite taken aback and unsure about how they should write or speak their answer (it is a ‘bot’ after all). Thankfully they warmed up to it.

Likewise, if the bot managed to accurately capture the data the users were rather impressed and felt more confident interacting with it. However, as soon as something went wrong they would get rather disappointed.


Helping humans be more human and less machine.

Humans are so ‘technologically’ advanced that there are things we can do that simply can’t be recreated by any machine, (any time in the near or distant future anyway). At Jade, we see it as our role to help businesses and people get the most out of AI rather than put people out of jobs. We free people up to tackle advanced tasks or problems that require more knowledge, emotion, and societal context than what AI can currently do. 

On the reverse side, we let machines be more machine-like, ensuring they stick to heavier computational tasks like processing huge amounts of data to find insights and patterns. This approach has significant economic benefit for a business and some other personal benefits for employees who can put up their feet or hit the golf course on a Friday afternoon because all their work is being done by a machine (if only this was the case).


Considerations for designing text, audio, and avatar-based digital employees

One of the great things about taking a conversation approach to digital employees (or undertaking a chatbot conversational transformation) is that you can have your conversations across a range of mediums. You can speak to an avatar (virtual human), transfer the conversation to a phone call as you head out the door, then take it to a text conversation as you enter a place where you aren't free to talk. But do you need to approach the various mediums differently?

Text based digital employees come with relatively few challenges because it's the medium users are most familiar with. The biggest considerations are keeping messages brief, and ensuring you have appropriate pauses throughout the conversation. 

Audio based employees come with a few more challenges. Firstly, it’s training your digital employee to understand speech. This uses technology called speech-to-text, and often some interesting answers emerge. In designing the conversation you also have to take into account what sounds natural when spoken. 

Grammar is important for helping the conversation flow, using punctuation to provide appropriate pauses. Pronunciation of foreign words can also be a challenge. You can often get around this using phonetic pronunciation. 

Response time is one of the bigger issues for avatar-based digital employees. This is no surprise due to all the processes going on under the surface. Considering all the emotional cues the avatar is picking up and responding to, as well as the conversation that it’s comprehending and then articulating, it’s understandable that the responses are longer than they should be. But this is no excuse. People want near-immediate answers and can get frustrated if they don’t get them. Investing in the right technology stack can pay dividends here.

There are also avatar-specific challenges, like “uncanny valley”. This is a technical term which basically means it feels like it’s almost real, but not quite right and you can’t place your finger on what the problem is. People’s visual expectations are different. They'll be looking for more social cues, emotions, and small movements humans often do unconsciously during conversations.


Businesses on the fence about adopting AI

In the chatbot space, businesses in general have not been ready to integrate AI-powered chatbots into their current systems. As such, there’s an abundant supply of FAQ bots, rather than the intelligent digital employees we offer. Everyone seems to want an intelligent chatbot, but don't always fully understand the current capabilities of the solutions. It’s actually relatively easy to integrate with systems. The main problem is access and security, which Jade has considerable experience managing.

There's also a business mindset that AI products can come ready-made and good to go. 

If they've provided pools of data, then sure that may be the case. But for things like digital employees, their use cases can be very specific to that organisation. It can be easy to get a base version up and running but they also need to invest time into getting users to test and break it. This ensures the bot has more data to learn from.


Parting thoughts on AI and digital employees

If someone is selling you an out-of-the-box AI system... run! The days of buying software out of a box on the shelf of a computer store are long gone. AI needs to be gently introduced into a company, within the right framework. There’s so much that could go wrong if it’s not done properly, and so much that can go right if it’s done well.

So if you’re ready to talk AI, we’d love to pick your brains and have a chat.



Get in touch!



Back to all posts >

Comments

Other posts