Reinventing Retail with Artificial Intelligence
My name is Declan Hoare. I’m Co-founder and Chief Technical Officer of Ludex, a cognitive computing company, and IBM Ecosystem Partner. We’ve been developing applications using Watson services, and we’re delighted to deliver a conversational chatbot for use by global airlines, and the social customer care dashboard for social media like Twitter. I know this is a complicated subject I’m going to cover so there’ll be a Q&A that I’ll open up to at the end if anybody has any questions.
Watson isn’t [inaudible 00:00:34] after Dr. Watson, Sherlock Holmes’ sidekick, it’s named after this gentleman, Thomas J. Watson. We’ve come a long way from the time when people thought there would only be five computers in the world. There are a lot more now. We can also access Watson through the cloud as a service. I’d like to talk now about how Watson came to be, its capabilities, and the concepts that underlie cognitive computing in general. Excuse me for having my printed out notes, I thought I’d go old school since we’re talking about Artificial Intelligence.
Computing from the 1800s to today has gone through three major areas: tabulating systems, programmable systems, and now cognitive. The computer was born not for entertainment or for email, but to solve a serious number crunching crisis. By 1880, the U.S. population had grown so large that it took more than seven years to tabulate the census. In 1890, a man called Herman Hollerith designed a punch card system to tabulate the census, accomplishing the task in just three years and saving the U.S. government $5 million. He established a company that went on to become IBM.
Three capabilities differentiate cognitive systems from traditional program systems.
How does cognition work? What drives the power of cognition is a common cognitive framework that all humans use to inform their decisions. We all observe, we interpret, evaluate, and decide, including the most advanced experts in the world. All experts use the same cognitive framework to achieve mastery over whatever their given subject is. Through hands-on experience and meaningful feedback over time, experts learn from a body of knowledge and apply it to make decisions. What changes is the corpus of knowledge. Now, experts are the lifeblood of any organization, but nowadays even the top experts can’t keep up. What if a system can help us reason over the universe of data, information, and knowledge, that could inform the decisions and work that we do every day?
Watson scales expertise to expand what’s possible in the workplace. Watson is ushering in a new era of computing. It’s fundamentally different from the conventional programmable computer systems that we’re used to. It mirrors the same cognitive process that we use every day to understand the world around us. With its ability to learn and navigate the language and protocols of specific industries and communicate in natural language, Watson is revolutionizing the way we make decisions, become experts, and share expertise at scale.
How are they different? Three capabilities differentiate cognitive systems from traditional systems: Reasoning, they reason. They understand underlying hypotheses. They form hypotheses. They infer and extract concepts, very much like a human. They never stop learning. That’s one of the key things with Artificial Intelligence and Watson, is that the more they’re used, the more they learn. Advancing with each new piece of information, interaction, and outcome over time, they develop expertise. And finally, understanding. They can understand like humans do, allowing them to interact with humans and enhance human performance.
Let’s talk about chatbots. It’s important to remember that the goal of most chatbots is not to match the capabilities of a human, to pass the Turing test which some of you would know about, but to help people achieve specific goals without needing another human being. Their goal is simply to help humans get answers, and buy stuff in a way that’s better than installing yet another app or figuring out yet another website. For many businesses, a chatbot system that automatically handles even 30% of customer requests, is still a huge cost saver.
This is the social customer care dashboard that you saw yesterday. It allows businesses to monitor their social channels and use Watson to analyze messages, detect their emotional tone like anger, sadness, disgust, happiness, and the person’s intent, the thing that they’re trying to do, what it is they’re talking about. With this, on top of that, on the left-hand side of the screen, you can, sorry, you can’t see it on this one, you can do Personality Insights. Personality Insights is a very complicated thing, but it actually follows on from what the lady from Google was speaking about in terms of, so I’d like to talk about that a little bit even though it wasn’t demoed yesterday.
What’s Personality Insight? It’s a Watson service. It’s based on the psychology of language in combination with data analytic algorithms. The characteristics are described in terms of three models, the first and most important is the Big Five, which as you can see are agreeableness, conscientiousness, extraversion, emotional range, and openness. Now, agreeableness is a person’s tendency to be compassionate and cooperative. Conscientiousness, their tendency to act in an organized or thoughtful way. Extraversion, their tendency to seek stimulation in the companies of others. Their emotional range is the extent to which their emotions are sensitive to their environment. Finally, openness is the extent to which they’re open to experiencing a variety of activities.
Now, each of these categories has subcategories that you can see up here, that allows you to match them up with the consumer needs and values. I know that’s a little confusing. Let’s take an example. Agreeableness, so people who score highly in agreeableness would exhibit signs of altruism, trust, sympathy. Those who score low would be shown as self-focused, cautious of others, and hard-hearted. Now, why does that matter? Combined with the other two models, Needs and Values, which you can see on the right and the middle there, Needs specifically describes from a marketing’s perspective the aspects of a product that are likely to resonate with the author of the social media posts, and Values, which describes from a psychology’s perspective the motivating factors that influence the author’s decision-making.
On this particular individual, I have somebody famous off the internet that I’ve done it on. If you look up here, you can see that an awful lot of that person’s scores are very, very high in terms of openness, conscientiousness, agreeableness, harmony, and love. The reason for that is …
What can that do for you? The Personality Insight services enables you to drive insights from social media, enterprise data or any other digital communications, email, text messages, tweets, forum posts. The service can then automatically infer a portrait of individuals that reflects a personality characteristics. It can determine their individual consumer preferences, which indicates their likelihood to prefer various products and services or activities. On a business level, Personality Insights can help businesses learn their client preferences and improve their satisfaction and strengthen the relations. But these insights can also be used to improve their client acquisition retention and engagement, and also to give a very, very guided personal engagements and interactions to better tailor your product’s services or whatever else for individual customers.
What are the practical applications then? Because I know that’s what you want to know about. IBM have recently conducted some research on this very, very subject, where they got a load of psychologists to go out into the field, contact people whom they had done Personality Insights on with big Twitter feeds or whatever, and give them a psychological analysis, which is basically a questionnaire that they’ve filled in. When they compared the responses to the Personality Insights that they had run on it, they found some very interesting results.
I’d like to share a couple of them with you. People who measure highly for modesty, openness, and friendliness, are also more likely to retweet information when asked by an organization. The second one is people with high openness and the low emotional range as inferred from their language and social media, they respond favorably to clicking an advertisement link or following an account. Now, targeting the top 10% of this group in an ad campaign increased the click rate from 6.8 to 11.3%, and the follow rate from 4.7 to 8.8%. Now, that’s an increase of almost 80%, specifically by using the Personality Insights to target who you’re targeting.