Wednesday, June 29, 2016

Creating Personalized Practice for Students


For a few decades now, educators have suggested that computers would vastly improve our ability to teach students.  The assumption has been that with computers we would be able to transport students to places they could not go on their own, allow them to communicate with people around the globe, and get more personalized instruction.

So far, the influence of computers in education has been mixed.  Computers clearly allow access to a much larger library of materials than students would were able to get before the internet.  In addition, students can view interesting videos that bring more content than the filmstrips and videos that were the norm in the past. 

However, computers have had a downside as well.  The internet provides a lot of distraction that leads students to try to multitask in ways that hamper learning.  Video lectures (like MOOCs) have not yet lived up to the hype.  It is just hard to watch a lecture on a screen.   And there have been few successful methods for personalizing instruction to individual student needs.

This issue of personalized practice was explored in a paper by Robert Lindsey, Jeffery Shroyer, Harold Pashler, and Michael Mozer in the March, 2014 issue of Psychological Science. 

They compared three methods for helping eighth-grade students to learn Spanish vocabulary.  For each of these methods, all students studied for the same amount of time. 

One group of students just practiced the most recent lesson each week for 10 weeks.  This process is similar to what many students do now when their weekly quizzes focus on the most recent material.  A second group got a generic kind of practice that spaced practice out over weeks in which they were given the chance to study both material from the current lesson and from the previous lesson. 

A third group used a more complex technique to guess what material each student was most likely to be just about to forget.  This technique used a student’s performance each week to estimate their overall ability to remember vocabulary, the difficulty of particular items (judged by how well students in general were able to learn those items) and how often they had seen those items previously.  Using this technique, students spent most of their time each week on the new items from the current lesson, but then received a smattering of older items to help them with older information that they would be likely to forget.

Students got a test at the end of the semester and another test after a four-week break as they started the next semester.  On both tests, students given the personalized training regimen did much better than those who either studied just the words from the current week or those who studied the words from the previous week and those from the current week.  The biggest benefit for the personalized training came for items that were learned early in the semester.  Students given this training were much less likely to forget those items than those given the other two techniques.

This finding suggests that when students are learning information that requires a lot of practice, computers may be used to individualize practice for the needs of particular students.  Of course, computers are just enhancing a piece of the learning here.  Picking up a new language requires learning vocabulary, but it also requires learning to actually communicate.  This technique helps with memorization, but not with the use of the new language.

One concern that I have with this technique, though, is that students need to learn material, but they also need to learn to study effectively.  That is, students need to learn that the reason that they succeed with this new technique is because it finds difficult items from the past that they are just about to forget and gets them to study those items.  These students will not always have a computer in front of them helping them to figure out what material to study.  So, they need to learn to apply some of these strategies on their own in order to maximize their ability to learn in the future.   

Wednesday, June 22, 2016

Studying Learning Using Video Games


A subtheme in this blog has been the influence of video game play on people’s behavior.  There is an ongoing debate about both the potential negative impact of playing games (such as increases in aggression or lower grades) as well as positive influences (like an increase in prosocial behavior after playing games with positive messages and an increase in speed of decision making).

Video games themselves can also be a great testing ground for theories of learning.  That was the approach taken in a fascinating paper in the February, 2014 issue of Psychological Science by Tom Stafford and Michael Dewar.

They were interested in testing a fascinating theory about learning new skills.  The idea is that any time you are learning something new, there is a tradeoff between exploiting past knowledge and exploring new strategies.  The proposal was that people who succeed best in learning new skills start by exploring a range of strategies.  This exploration may cause them to do well when they hit on a potentially good strategy and to do poorly when they try something that does not work well. Then, when people find a good strategy, they should exploit it and keep improving their skill.  Those who settle on a strategy too quickly (and so they start exploiting early) may end up picking one that is far from optimal, and so they will do more poorly in the long run.

They worked with a game designer to create an on-line game called Axon.  The game involves making simple selections of targets on a screen.  Some of the dots are colored, and the colored dots have different influences.  Some colors are better to hit than others.  So, there is some strategy, but the strategies are not complex and are mostly focused on which color dots should be selected. 

During the data collection period, the game was played over 3.5 million times by over 850,000 players.  Most of the analyses were done on the 45,672 individuals who played the game 10 or more times.

Before addressing the key experimental question, it was important for the researchers to make sure that the data gave sensible results.  So, they started by verifying that two classic phenomena from learning studies occurred in this data set.  The first was the observation that performance improves with practice.  Looking at those people who played at least 10 games, it was clear that the more games people played, the better they got. 

Second, it is well known that when people space their practice out over time, their performance is better than when it is concentrated together.  There are several reasons for the improvement for practice that is spaced out over time.  One is that spaced practice allows information being learned to become associated with many different contexts. The second is that sleep improves skilled performance, and the more times people play and the more time that elapses between attempts, the more likely that people will sleep between attempts at the game.

Consistent with the prior work on this topic, people who concentrated their game play together did worse at the game than those who spread their play out over time.

Because the game yields sensible data, the researchers could look at the more complex question relating early strategies to later performance.  The researchers looked at the degree of variability in the scores over the first five games people played.  The more variable the scores, the more strategies people were likely to have tried early on.  They looked at how this variability was related to the high score people achieved on the next five games they played. 

The more variable people’s scores in the first 5 games, the higher their best score tended to be in the next five games.  That is, exploring a range of initial strategies allowed participants to do better than selecting a single strategy early on and sticking with it.  

There are two interesting things about this project.

First, it is a great example of using the internet to look at learning in a large number of people.  Most research studies done at universities involve a small number of participants who perform a small number of trials on a task.  This game is still fairly simple, but it allowed a large number of people to play it.

Second, these findings suggest that it is valuable to try a number of different strategies in any task before settling on one.  People who did this naturally in the game were more successful than those who did not.  An interesting follow up study would explore whether people who are instructed to try different strategies also do better than those who are instructed to select a strategy and stick with it.  That would disentangle whether these results reflect that people who are better at video games are those who also try more strategies, or whether anyone can improve by trying more strategies.