Niels Taatgen

HomeResearchOverview Skill Acquisition Multi-tasking Time perception Learning from instructions Transfer: PRIMsPublicationsPersonal


Learning from instructions

If we have a good theory about learning, can this also help us to improve learning situations? Anderson's work on algebra models has shown that they can serve as a fruitful basis for cognitive tutors, programs that keep track of the student's knowledge, and select new exercises and feedback accordingly.

My own work, funded by NASA and the Air Force, concerns modeling how people learn from instrustions and eventually derive robust knowledge that leads to flexible performance. That is, even if the task is different, or interrupted, or otherwise changed, people can easily adapt and do the right thing. The domain we study this is programming the flight management system (FMS) of a Boeing 777.

The FMS is a computer system that, when properly set up, can fly the plane on its own with the exception of take off, and in most cases, landing. Programming the FMS is part of the pilot's training program when they first have to fly on airplanes that have an FMS. In the current training program pilots have memorize procedures, which consist of lists of steps that you have to do. This type of learning, however, is not very successful, because pilots have a hard time memorizing infrequent procedures, and they are furthermore often unable to apply procedures to slightly different circumstances. In other words, their skills are neither flexible nor robust.

In our models of skill acquisition we found that a better representation of instructions is to not just list the actions, but also tell you under what circumstances an action has to be carried out, and what its intended effect is. That idea can be carried over to how we instruct people: if we extend instructions by specifying this additional information, they might learn faster and more robustly. As a small example: the first step in one of the procedures we looked at is:

1. Press the LEGS key

Instead of just listing the action, we can also specify why you might want to press the LEGS key:

If you want to go to the LEGS page, press the LEGS key, which will bring you to the LEGS page.

Now suppose you are in the following situation (this is part of the experiment):

This would be a case where the FMS is already on the LEGS page (if you would be in the experiment you would recognize the "LEGS" text on top of the display), and it would be unnecessary to press the LEGS key. In experiments in which we gave people both the standard list instructions, and instructions that we augmented to take context into account, we found that people indeed learned faster, and were able to better solve new problems when they received instructions with context. And, they pressed the Legs key significantly less often!

In our latest experiments we have been looking at eye-movements while doing the FMS task. One concern we had about the extended instructions (which we call the context instructions) is that people might skip essential verifications steps. One of these steps is that before people finalize the procedure, they have to check the navigational display (which is the top-right part of the experimental interface above) to see if the modifications they made are correct. It turns out that people that received the list instructions made this check in only 46% of the trials, while people who received the extended context instructions made the check in 71% of the cases. Click here for a sample playback movie (requires Quicktime).


This project is funded by the Air Force Organization for Scientific Research

Key references

Taatgen, N. A., Huss, D., Dickison, D. & Anderson, J. R. (2008). The acquisition of robust and flexible cognitive skills. Journal of Experimental Psychology: General, 137(3), 548-565. (pdf)

Taatgen, N.A., Huss, D. & Anderson, J.R. (2006). How Cognitive Models can Inform the Design of Instructions. Proceedings of the seventh international conference on cognitive modeling (pp. 304-309). Trieste, Italy: Edizioni Goliardiche. (pdf)