Computers, creativity and learning

Help - could a computer think?

I'm working on some materials for primary school teachers to help them teach children about computers. Can you help? I am now focusing on artificial intelligence after have some fascinating philosophical discussions with kids about whether machines can think. I wanted to summarise what is known about AI in a non-confusing way without going too deep into the raging debate about our Robot Overlords or indeed giving children nightmares. Below is my attempt at an explanation of the state of the art of AI. 

• If you're a teacher, is there anything else you need to know about this topic, or anything which is not clear?
• If you're an AI researcher (or an AI or that matter), is there a key concept I have missed, or anything which is deeply misleading?

Could a computer think?

Most people would agree that the computers we use every day don’t think for themselves in the way that people do– they follow instructions which people have given them. Sometimes it seems like the computer or phone is more clever than it really is, especially if you are using a program which seems to understand what you say. Computers are good at calculating, searching huge amounts of information very quickly and accurately and sometimes making decisions based on the information.  Humans are not so good at this sort of task, but we’re usually much better than computers at working out answers to new problems, being creative and understanding feelings.

Scientists and engineers who research artificial intelligence have been trying to make intelligent machines since the 1950s. Because of their work, your phone can recognise what you say to it, satellite navigation programs (“Sat Nav”) can speak directions out loud, robots (such as the Mars Rover) can navigate for themselves, and even the best human chess players can be beaten by a computer. Although these programs are impressive, they can’t do all of the many amazing things which even young children can do. Activities which we take for granted, such as walking, understanding how your friend feels or learning new languages turn out to be surprisingly hard to program into computers.

Could computers in the future think for themselves? And if they did, would they be friendly to humans, or would they want to wipe us out? Nobody really knows yet, and experts disagree about it. But it is worth thinking hard about it today because it may be possible to design intelligent computers for the future which can help us to make Planet Earth a fairer, safer and more interesting place to be.

So-called digital natives can't solve problems with technology

Remember the "digital native" hype from the early 2000s? There was a lot of discussion about how there was a new generation of children growing up were born with access to technology, and that their technological prowess  would be such that traditional education would need to reform to accommodate it. Research evidence is now growing to confirm that the superior skills of digital natives are in fact not a reality. So you can feel quite smug if you rolled your eyes every time someone mentioned "digital natives" since 2001.

A report from Change the Equation -a US non for profit group which advocates STEM education- recently reported that 58% of Americans aged between 16 and 34 have poor skills when it comes to solving problems using technology. Let's take a look at this in more depth.

What sort of research is this?

The recently published study is a re-analysis of a previously published data set from the OECD Skills Outlook Survey in 2013. The original study contained data on literacy, numeracy and technological competency from 166000 adults in 24 OECD countries. The re-analysis focused on the 5000 US respondents between 16 and 64,and looked more closely at the relationship between technology skills and income. The researchers used a quantitative analysis method known as regression, which examines the relationship between variables. Regression does not tell us whether one variable causes another; related variables may share an underlying variable, for example.

Assessment of technology problem solving skills came from a set of computer tasks which respondents were asked to complete. At level 1 (the easiest), tasks involve the use of common software such as email programs or web browsers to plan and use information. For example a level 1 task is "Categorise a small number of messages in an e-mail application in existing folders according to a single criterion.” Level 3, the most proficient level, requires participants to "Set goals and monitor progress, plan, acquire and evaluate information and use information". An example task involves "multiple applications, a large number of steps, a built-in impasse, and the discovery and use of ad hoc commands in a novel environment", all in the pursuit of the exciting goal of booking a meeting room.(Yes, really).

What are the main findings?

58% of 16-34 year olds in the US cohort had poor skills in solving problems using technology.People at the highest skill level earn more than twice as much as those in the lowest skill level, even when other personal characteristics such as education, age, race, gender, literacy and numeracy variables are controlled. 88% of those with low skills believed that their lack of technology skills would not hold them back in job seeking, promotions or pay rises, which is contrary to employers' views.

For comparison, a quick look at the original England/Northern Ireland data for 16-24 year olds shows that 54% were assessed at level 1 or lower. No data was available for Scotland.

So what?

The workforce in general (and even young people who are generally thought to be more proficient in technology use) have difficulty using technology to perform tasks which require an element of problem solving. In the US, lower levels of technology skill is associated with substantially lower income. We should be careful not to infer from this that teaching kids more technology skills will in itself increase their income as adults, but it is suggestive. More broadly, I suppose it might mean the potential increases in productivity in the workplace which could be brought about technology are not happening in practice.

We can't assume that young people have magically absorbed the ability to use technology in support of complex tasks just because they spend a lot of time in front of a screen. We can't rely on the non-pedagogy of osmosis, as my colleague Rory would say. This means classroom teaching (from pre-school to university) should focus on solving problems with technology (rather than focusing on software functionality, say, showing kids how to use the "bold" function in a word processor).

But I think we can do better than introducing kids to technology with mundane administrative tasks, as were assessed in the original study. This is related to the computational thinking agenda which is becoming dominant in computer science education circles internationally.  Computational thinking refers to the theoretical roots of the intellectual discipline of computer science(it is not just about optimising the process of jam sandwich production although you could be forgiven for thinking so). It doesn't strictly speaking require a computer, but in practice is often taught with technology tools. Instead of teaching young people how to harness the power of modern computation to book a meeting room, let's give them the intellectual and technological tools to send robots to other planets, visualise the human genome or process data to help us decipher the fundamental nature of the universe.

 

References

Read this for a cute poster :http://changetheequation.org/sites/default/files/CTE_Technology_Skills_Infographic.pdf

Full report: http://changetheequation.org/sites/default/files/CTE_VitalSigns_TechBrief.pdf

Methodology: http://changetheequation.org/does-not-compute-background-and-methodology

Original PIAAC report: http://piaacgateway.com/what-is-piaac/

Critical review article about Digital Natives: http://onlinelibrary.wiley.com/doi/10.1111/j.1467-8535.2007.00793.x/abstract