Roger Schank – making better expert software

There are much better ways to build software which can make a useful contribution to experts’ lives, said Roger Schank, CEO of Socratic Arts. One way is to start with story telling

There are much better ways to build software which can make a useful contribution to experts’ lives, said Roger Schank, CEO of Socratic Arts, and perhaps a good place to start is story telling.

He was speaking at the Oct 29 Athens conference “Software for Domain Experts” download the full report

Mr Schank’s company Socratic Arts develops performance improvement solutions for companies. He is also CEO of XTOL Corp, a company which offers technology career programs. He is Professor Emeritus of Computer Science, Psychology and Education at Northwestern University, Illinois (USA), specialising in artificial intelligence, learning theory, and cognitive psychology, and a former chief education officer of Carnegie Mellon University’s Silicon Valley campus.

Mr Schank presented a video of Dimitris Lyras, the chairman of the conference, telling a story from several decades earlier of how his shipping company had a ship delayed in a port in the Caribbean.

The crew went fishing for barracudas, which were fried, roasted, boiled and eaten onboard. Unfortunately the barracudas were poisonous because of the plankton they eat. The poison can be lethal. None of the crew died but some had to go to hospital onshore and be sent home, and the ship left the port without an electrician onboard.

This story is irrelevant to nearly everybody, Mr Schank said.

But if you were running a ship which was entering the Caribbean, the ship was going to be there for a while, and someone takes out a fishing rod, it would be a good story to know.

Meanwhile modern computer software, and the systems we use to run our organisations, is nowhere near being able to provide this information, or this story, at the right time. People are usually asked to go on a course or read a manual to learn what to do.

It is not hard to find similar examples. Mr Schank presented a video of a US solider who had been asked to help set up a school in Afghanistan, working with someone locally. The soldier can follow orders but has no idea how to do it. But there might be someone else in the army who has done it before. The organisational challenge for the US Army should be connecting this soldier with stories from the soldier who has done it before, Mr Schank said.

A further example was of a ship going through the Suez Canal which had a fire in the boiler room. The chief engineer started to shut down the engine, as per the instructions in the manual.

The captain told him that was a crazy thing to do. The ship would be hit by the ship just behind it in the canal. If it entered port, the authorities might demand a bribe or try to steal the cargo.

The point here is that we are still dependent on having the right person standing behind us at the right time to provide us with relevant advice. Software ought to be able to bring us the right story at the right time, but in 2015, it can’t, Mr Schank said.

Yet as people, all of us are experts at this. “That’s what we do,” Mr Schank said. “We say something, and someone says, ‘that’s what happened to me’”.

So bringing up the right story at the right time could be considered a reasonable definition of what ‘human intelligence’ actually means.

Artificial intelligence

Mr Schank is recognised as a world expert in the field of ‘artificial intelligence’. Yet he has become very disillusioned by the field, he said.

Consider IBM’s “Watson”, a computer system developed by IBM to answer questions on the US quiz show ‘Jeopardy’, and won a $1m prize, competing against former winners. Watson is claimed to be an example that ‘artificial intelligence’ could now be better than human intelligence.

But there is also an argument that Watson is nothing more than a search engine, Mr Schank said.

As an example, one of the questions in the Jeopardy show was, “It was the anatomical oddity of U.S. Gymnast George Eyser, who won a gold medal on the parallel bars in 1904.” Watson gave the answer ‘a leg’ which was counted as an incorrect answer.

What Watson probably did, Mr Schank said, is look at the Wikipedia page for George Eyser, which said “Eyser competed with a wooden prosthesis for a left leg”. Watson could understand that the answer had something to do with a leg, but all of its artificial intelligence power was not able to connect the phrases “anatomical oddity” and “wooden prosthesis”.

It is important not to confuse statistical analysis and pattern generation, which computers can do better and faster than human, with the deeper pattern generation which humans can do, and this is probably what ‘intelligence’ is, he said.

You should also be wary of technology companies making marketing claims that their products are ‘intelligent’ when they are nothing of the sort, he said.

One computer training company claims to offer the ‘future of study’ by automatically creating study aids using content, he said. On closer inspection the software seem to just find content on Wikipedia relevant to what you want to study, serves up some text with some words deleted, and asks students to fill in the blanks, he said.

As an Artificial Intelligence researcher, “I thought we were trying to figure out how the mind works and making cool stuff,” he said. “This is not cool stuff or figuring out how the mind works.”

There seems to be a growing belief in society that expertise is something you can get from text books, not something you gain from experience, Mr Schank said.

But you don’t need a text book to “do what I’m doing right now, having an idea and talking about it,” he said.

Expertise is really shared through stories. “When people are telling stories, they are having an emotional experience,” he said. “Emotions are the real stuff. That’s when you talk to someone, you know how they’re feeling because you see it in their eyes. This is what life is about. Watson doesn’t have any emotions. Watson isn’t excited about anything.”

 

Video for assessing people

Mr Schank demonstrated software built by his company Socratic Arts for assessing people, which could work just as well for finding professionals (for example a doctor), finding a life partner (a date), or finding an employee for your company.

People who would like to be considered as candidates (as your doctor, date or employee) record videos about themselves answering questions about important topics, such as how they dealt with a tricky situation (a doctor) or why they want to be a programmer (a potential employee).

Watching someone on a video, we can usually make a judgement about them in a few seconds. And whether our judgement is good or bad, it will probably be the same judgement we would have made after a 30 minute face to face interview, which is far more time consuming.

There is still light computer processing involved – the computer can spot patterns in the type of rankings different videos get, and use this to (for example) bring you someone very different to the last person you saw, or very similar.

Compare this to other software tools on the market which try to use computer algorithms to suggest you a life partner who would be a good ‘match’, but have only text answers to go on, such as ‘I like to laugh a lot’.

In today’s society, the best way we can find a professional to help us (for example a psychotherapist or doctor), is by looking at online profiles people have written about themselves on websites, which are not usually very helpful.

“My premise here is that this kind of thing could change the world if and only if we get the stories,” he said. “We need people sitting in front of a camera and saying what they think that’s relevant.”

Mr Schank told a story about his father, who “”never accomplished much in his life.”

Mr Schank was a graduate of Columbia Law School, and was on a work experience placement together with a graduate of the less elite New York Law School. “I made sure to give him all the junk work,” Mr Schank’s father would say.

The graduate of New York Law School went on to be Chief Justice of the State of New York.

You can see why a father might want to tell this story to his son if his son was being a bit cocky, as a sign that hard work and humility might take you further than arrogance. It is also a story which would be good to hear at the right time.

“We don’t always have the right guy to tell you something at the right moment, but you could,” he said.

Too much complexity

Another problem with most enterprise software is that it is too complex, he said. Consider that most organisation education software products allow you to upload as many PowerPoints, videos and documents if you want.

But the best education software might do none of this – it would just give students one task to do, and once a (human) teacher agrees it is completed, the teacher will put you onto the next task, and provide assistance as required.

This software might not be commercially successful though. “Teachers prefer talking,” he said.

Also the educational establishment might not like this idea, because educational professionals often don’t understand the concept of goals, he said.

“But goals are the driving force for absolute all knowledge. There’s nothing we do which doesn’t have a goal. Goal is driving every single aspect of your day.”

But in the educational world, the emphasis is usually on ‘creating courses’ or trying to cover topics, he said.

Similarly in the Artificial Intelligence world, many companies talked about trying to build ‘general problem solvers’.

Mr Schank always argued that this was impossible. “People who are good at solving problems in one area not good at fixing problems in another area. The idea that there must be such a thing as general problem solvers, it is simply wrong,” he said.

Leave a Reply