Co.Design

IBM's Next Big Thing: Psychic Twitter Bots

Using the power of the Watson supercomputer, your bank could soon be stalking you on Twitter. And that might not be a bad thing.

Using some of the same technologies that allowed the Watson natural language supercomputer to conquer Jeopardy, IBM's next step: Psychic artificial intelligences that read your Twitter feed and can tell when you're about to have a baby, get married, buy a house, or move across the country, and even tell you how these major life events make you feel, then approach you about them accordingly.

Last week, IBM announced a new consulting practice called IBM Interactive Experience dedicated to better fusing business strategy, data, and design. As part of the announcement, IBM revealed that the company had been working on perfecting a couple of tools that are offshoots of the same computer research into natural language that led to Watson.

Those tools? Life-event detection and psycholinguistic analysis. The idea is pretty simple, and it's all a part of IBM's initiative to help companies design the customer experience of tomorrow. What if a computer using linguistic analysis could read customers' social media profiles to better understand who they actually are, then use that data to figure out a better way to approach them, and predict the services or products they'll need in the future?

For example, let's say that you tweet that you've gotten a job offer to move to San Francisco. Using IBM's linguistic analysis technologies, your bank would analyze your Twitter feed and not only tailor services it could offer you ahead of the move--for example, helping you move your account to another branch, or offering you a loan for a new house--but also judge your psychological profile based upon the tone of your messages about the move, giving advice to your bank's representatives about the best way to contact you.

A company using IBM's tools wouldn't even need to already know your Twitter handle to figure out how to approach you. These technologies are capable of scanning publicly available social media streams like Twitter, Facebook, Instagram, LinkedIn and more, and then use linguistic analysis to try to infer which social media accounts belong to which customers.

"Twitter handles don't usually correspond to customer records," says Rakesh Mohan, IBM's business services research director. "So we need to make that linkage." IBM's technology can scan a publicly visible social media account for basic profile information, locations, and other snippets of information that would allow IBM's clients to construct a digital fingerprint. That can then be compared to customer records to make what Mohan calls a "probablistic inference" that a given Twitter handle, say, corresponds to this or that real-life person.

For example, let's say you go on a road trip to Canada from your home in New York, and decide to tweet about it. With the information it already has about you--age, address, transactions, and so on--your bank could use IBM's technologies to figure out your Twitter handle based upon what restaurants you checked in at, Instagrams you took, towns you mentioned passing through, and so on. Using these techniques, a company could identify every social media account one of its customers has.

And it doesn't stop there. Once a company knows a social media account is yours, IBM's tools could allow companies to know when you're thinking of going on vacation, making a big purchase, strapped for cash, or more. It can even predict major life-events: if you changed your Facebook status to "Married" a year ago, for example, a company might infer that it was about time to start approaching you about products and services for your first child.

From a consumer point of view, it all sounds a little creepy at first, but IBM is quick to point out that they are not building a database linking social media users with real-life people.

Rather, IBM wants to offer tools that will allow companies to understand their customers better by contextualizing through public social media data what is happening in their lives. If customers don't want companies to contextualize them in this way, they can simply change their social media privacy settings, and IBM is also looking into developing a more global opt-out feature.

The ultimate hope is that these tools will help companies be less spammy and tone-deaf in the way they approach customers.

"No one actually wants to bombard customers with irrelevant messaging, and likewise, no one wants to be spammed," says Paul Papas, managing partner of the IBM Interactive Experience. "Using the deeper understanding that psycholinguistic analysis can bring to the table, the experience can be more personalized. On the customer side, consumers won't be bombarded with messages that don't make sense, and on a company's side, they aren't wasting their time or money to create dissatisfaction. It's win-win."

Ultimately, though, what companies end up doing with these tools is unknown. The power of Watson-like computer language technologies could either lead to a more pleasant world of customer service--one in which a company can tell the moment you call it whether you're an introvert or extrovert, or happy or depressed, and channel you to the service rep who best matches your personality right from the get-go, or where you are only approached by a company with a product when you already think you might want it. But it could also potentially lead to a more intrusive world.

Like most powerful tools, whether IBM's psychic Twitter bots end up being bane or boon will end up having a lot to do with what companies choose to do with it.

[Image: Psychedelic via Shutterstock]

Add New Comment

2 Comments

  • While this looks technically savvy, the marketing insight doesn't appear to be very illuminating. The insight? The married group is more inclined to long term, fixed rate mortgages and the singles are inclined toward short term adjustable mortgages. Anyone could have "predicted" that without analytics. A more counter-intuitive outcome would be more persuasive.

  • At first when I read "... might not be a bad thing", I immediately thought of identity theft and protecting the consumer. Would be nice to have this elaborate effort focus on something like that too, especially if it is going to be used by banks.