Constraints are literally constraining or forcing the user in the way a thing is used. They limit how we can or perceive we can use a thing and come in many forms. They are almost as important as perceived afordances.
When designing an object you literally force the use of the thing to not physically work in some ways. Such as with don normans Lego bike example some parts just cant physically fit together and this constrains how we use it and where some parts must or must not go. You are given all the pieces to a small Lego bike and asked to assemble it. For digital screens its a bit harder as nothing is physical.
Logic helps us decide how to complete a task with a little problem solving. With the Lego bike logic would say that the rider faces forward not that it wont fit backwards but that they must be forward to operate it. Logic helps with websites so that users scroll down or look for more options and ways to use it when they believe the task can be completed from that page just not that screen.
This is the most important constraint. Culture dictates how many tasks and objects are to be done. Some cultural constraints are a small region like my wives home region of the navajo nation in Arizona the constraint about food is if there is a group of people you try to avoid taking the last piece of food especially break so you take a 1/2 portion. If there is 1 tortilla you tear it in half and put it back, the next person will probably tear that half into a quarter and put it back and the next will probably just take that small piece. This is cultural but only for a region of a few hundred square miles and I have never seen it any where else. A global culture is that numbers mean the same thing in any culture with writing. Such as 0 always means 0 globally from the USA, Europe, Africa and Asia. Some constraints are basically universal as that. Culture helps us decide how many things should be done and used. From a design standpoint you don’t want to be so innovative that there is no cultural constraints on it. Instead you would want to push those boundaries but still make peoples past knowledge be able to help in figuring out the thing.
Do make sure a designed set of constraints work you need to test it. Don Norman and Steve Krug both agree in separate books that user testing can be expensive, complete and very scientific. They also agree that user testing should be soon and often to avoid critical mistakes being perpetuated. They both say that user testing can be done with a handful of people just a few times and that will find most of the issues with the design of a website or other things. Cheap testing with small groups will work as long as you still do multiple rounds of testing through the course of development. They don’t need to be perfect examples of your target audience at first but just warm bodies. As testing progresses it would be better and better to get people who are closer to your target audience. At worst you will make a thing that is more usable by a wider group of people.
I still hear far too much dogmatism about what people really “want,” what they “believe,” or how they “really” behave, but I see very little data. It doesn’t take much data. My partner, Jakob Nielsen, has long argued that you can get these data at a discount: three to five people will give you enough for most purposes (Nielsen, 1993, 1994). But they need to be real people, doing real activities. Don’t speculate. Don’t argue. Observe. -Don Norman http://www.jnd.org/dn.mss/affordance_conv.html
Constraints are probably equally as important as afforances. Often limiting what a thing should be used for is more important than what it can actually be used for as perceived affordances is what we think a thing can be used for is more important and every possible use in the universe for the thing. Many constraints exist to keep the thing from being used in a way that will hurt itself, user or others. Some make sure certain steps are taken in order to use it a lockout, lock in and interlocks. Those are all constraints to how it will operate.