Education and Learning

8.    Education Learning and Communication

Education

You can’t teach someone without feedback. Well you can, but the results are somewhat unpredictable. That’s the whole point of the feedback. It tells you how much has been learnt, and how well, and what problems remain.

That is why schools have tests. They are often seen as just a means of judging how good a pupil is, and if he or she is a candidate for further education, or a scholarship. But really, they are just as much a marker of how well the education process is going. They tell the teacher what the student has learnt, and what still remains to be learnt.

If you have ever taken a golf lesson, you will know that the whole process is one of constant feedback. You hit a shot, and the pro tells you what was wrong with it (usually). He shows you how to make and adjustment, and you try again. This goes on (in my case at least) for a long time. Lots of feedback, and there is no short cut.

In fact feedback is perhaps the biggest difference between a real live teacher and a teaching machine. You can program a computer quite easily to impart information, but it is not so easy to give feedback. A lot of progress has been made in this direction since the first days of teaching machines, but it is still something that a human can usually do better. The problem is that the teacher needs to have an understanding of the individual student, and a model of his learning method, to best give ‘formative’ feedback. And more than that, enthusiastic encouragement is a benefit which is sometimes hard to take from a computer, however much they try.

Learning

Learning is what happens when you don’t have education (only joking). Animals and humans are all capable of learning by experience. Although some animal behaviour that looks so ‘intelligent’ that one would think it involves learning is, in fact, built in pre-programmed instinctive behaviour.

Courtship dances of fish and insects, pecking, nest building and egg rolling by birds are all examples of built in ‘motor programs’. These actions are innate, although they do allow for adjustment to deal with specific circumstances. And naturally, these adjustments require feedback from position sensors and vision.

Other motor programs, though, can be learned. In humans, activities such as walking, swimming and bicycle riding are learned behaviour. But once the learning process has been gone through repeatedly (a lot of times in some cases), it becomes almost automatic. Feedback is essential in the early learning stages, but becomes less important as the process becomes more automatic. Anyone who drives a car for a while is familiar with the ‘automatic pilot’ syndrome, where it is possible to drive back home from work and hardly be aware of making any conscious effort. Mind you, even then there is still feedback going on, otherwise it would be too easy to drive off the road.

In fact, driving involves a lot of feedback. Not only must you keep your eyes on the road, you also have to be aware of other traffic, pedestrians, obstacles, traffic lights, traffic signs and lane markings. Then you have to monitor the pressure of your foot on the accelerator, and sometimes the brake (and clutch if you insist on a manual gear change). When you watch speed up film of dense traffic, it sometimes amazes me that there are not a lot more accidents than there are.

Speech also requires feedback in the early stages, which is why deaf people have difficulty in learning to speak. But once speech is learnt, deafness later has little effect. The feedback is no longer needed.

Sometimes feedback like this can actually cause problems. Stutterers, for example, can sometimes benefit from an electronic device that prevents them hearing their own voice. It seems that the feedback can be part of the problem.

And indeed, stuttering can be induced by delaying the audio feedback to someone who is speaking. If you have ever tried to speak while listening via headphones to your own voice delayed by a second or two, you will know how difficult this makes it.

So all learning involves some kind of feedback to inform the learner about the effects of their actions. The other key aspect of learning is reinforcement, where desired behaviour is reinforced positively, and undesired negatively (gold stars or black marks – or, in my day, sometimes a ruler over your knuckles).

So feedback in learning is used both to adjust actions (such as motion, or pressure), and also to adjust the ‘world model’ that is being used to learn the behaviour. We learn, for instance, that some objects are hard, others are soft. Some are heavy, others light. We can only do this by taking action, and monitoring the feedback.

In the wider sense of the whole learning system, reinforcement itself is a kind of feedback. It provides positive or negative signals to the learning system to tell it when to repeat behaviour, and when to adjust it.

Some feedback is external - such as that provided by a teacher, and some is internal – such as neural signals fed back into the nervous system as part of learning movements.

If the feedback is reduced or hampered in some way, then learning is slower. Likewise, if it is enhanced or amplified, then learning can be improved. Good learning depends on good feedback; the clearer, and more specific is the feedback, the better the learning. Feedback also needs to be timely to be effective, delayed feedback is bad feedback. Practice makes perfect only if there is feedback. Practice without it leads to poorer performance and increased errors.

Feedback can also improve task activity that has already been learned. If a task required constant vigilance, it has been found that providing feedback to give knowledge of actions taken will improve the performance. It seems that the feedback stimulates the performer to better monitor his actions and so improve his vigilance.

Humans also learn how to behave (or not) by using feedback. Encouraging words, hugs, sweets and other desirable actions tend to lead to repeating the behaviour, while similar negative feedback tends to inhibit it. This does not mean, however that all behaviour is learnt. We are not ‘blank slates’ as Steven Pinker so ably demonstrates in his book of that name. Humans have a propensity for certain kinds of behaviour, but of course their subsequent path is hugely influenced by what they learn after birth.

Learning is actually a pretty tricky business. In Gregory Bateson’s book Steps towards an Ecology of Mind, he shows that people don’t just learn; they learn to learn. In fact, they also learn to learn to learn, and maybe that goes on further, depending on how smart you are. Some computer programs can learn, but getting them to learn to learn is a lot harder. When they start to do that, we may see really intelligent computers. But in any case, there has to be some feedback to make sure the learning process is doing any good.

Animal Learning

Animals also learn, though sometimes it can be difficult to distinguish learning from instinctive behaviour. For instance, Jackdaws learn what objects to use when making nests by trial and error (so using feedback). They will start off by grabbing just about anything, but learn that some things just don’t make good nests. Other birds, though, apparently have built in knowledge of what constitutes a good nest.

Bird song is an example of a combination of instinctive behaviour and learning. Birds that do not hear other birds have only a poor and limited repertoire of songs, but those exposed to bird society learn to develop a richer range.

And any pet owner knows that they can be trained to learn what kind of behaviour is acceptable, and what is not (well, some of them can).

Communication

Hey you, are you talking to me? Well obviously not, you are reading this, so in a sense I am talking to you. In the age of the media, communication is taking place in a variety of guises – printed media, TV, radio, web sites, podcasts, CDs, DVDs. And there are still the art galleries and concerts and books that have long been a source of communication.

Theoretical approaches to communication stem mainly from the 1940s, with the well known work of Claude Shannon. Shannon described communication channels between a transmitter and a receiver, and went on to explore the properties and limits of sending messages through these channels.

This kind of communication is obviously demonstrated by telephony or radio, but also by data signals between computers through modems or data networks.

Essentially though, this is a one way system. The message is sent and received, and that’s about it. Of course, there may be a message in reply, but that is separate and distinct from the first message.

Often though, communication is more complex than that. When we talk to someone, we are observing them while we talk, and we may be adjusting the tone or content of our ‘message’ by the way in which we see them reacting. So there is feedback here. We us the ‘signal’ from them as feedback to modify the way in which we are communicating.

Feedback is also used to overcome another feature of communication that Shannon explored – noise. In communication theory, noise is an unwanted signal that can interfere with the intended message and cause distortion and errors. We come across similar problems when talking to someone in a noisy room, or over a bad phone line. Often we overcome this by checking that the other person has heard us correctly. Did you get that?

This is also feedback, at the ‘message’ level. Similar techniques are used in data communications, by using additional error correcting bits to check the message, and even correct it if possible. This is an example of redundancy in communication, which we use all the time. Written text contains a large amount of redundancy, something that can easily be seen by omitting say vwls frm sm mssg. A technique that has received a large fillip from texting in mobile phones. It is also surprising that you can often read a piece of text by observing only the top half of a line.

 Modern mobile phones use huge amounts of digital signal processing to continuously monitor and correct the transmitted voice signal. Clever techniques are used to transmit a ‘training’ signal in each voice packet that contains known data, and can hence be used to adjust the parameters of processing algorithms that optimise the accuracy of the voice signal. This is all feedback, using information from the received signal to change the way in which the transmitter is sending the signal.

These are simple examples of communication (though the message may be far from simple). The modern world is rapidly becoming more diverse in its modes of communication. Newspapers, TV and radio are mainly a one way street. Someone writes an article, or transmits a program, and people read it or listen or watch. That model is changing. Of course, there has always been some feedback in this process – letters to the editor, viewers comments, but these have been limited in speed and quantity. Now, with the increased use of blogging, instant messageing, chat rooms and social networking the aim of the broadcaster is often to engage the end user as much as possible.

I am a regular listener to Podcasts from both TWIT (This Week in Tech) and Science Friday. As they are broadcast in the U.S., I am limited to listening to them from downloads, but both shows take live feedback from listeners (or viewers) via a variety of means – phone calls, Twitter, chat rooms, second life, email.