Why Is programming an unnatural activity?

For my P2PU course I have been looking at "Novice" programmers.  And in one of the papers we were asked to read Mark Guzdial asks:

“Why?” Is programming an unnatural activity? 

Could programming be made easier in a different form? 

Could programming be taught in a different way that makes learning easier? 

Or maybe we just have no idea how to actually measure what students know about programming. (1).

My main problem with the Guzdial paper (this was more my problem than a problem with the paper) is I felt it didn't provide enough details or specifics on "Why it is so hard to learn to Program?"  I need specifics and examples to get my head around things.  Roy Pea, was a great find and perhaps not surprisingly (for me at least) the Resnick article was very useful. 

Pea (et al) talked about three classes of bugs:

1. Parallelism Bugs
2. Intentionality Bugs
3. Egocentrism Bugs

Parrallelism Bugs

The Parallelism Bugs, is basically an "assumption of different lines in a program can be active or known by the computer at the same time or in parallel".  For example, look at this code:

If (Size == 10)
    print "Hello"
For Size in range(10):
    print Size
  

When High School students. in their second year of programming course, were asked what they thought the program would print 8 out of 15 predicted "Hello" would print after "10".

Intentionality Bugs

The Intentionality Bugs, is the idea in the child's mind that "the program has goals and knows or sees what will happen elsewhere in itself."

Egocentrism Bugs

The Egocentrism Bugs, stem from the belief that there "is more of their meaning for what they want to accomplish in the program than is actually present in the code."  Funny, I see these kinds of bugs all the time in my code and those of other experience programmers :) 

The Super Bug

He concludes that all these derive from the Super Bug:

The idea that there is a "hidden mind somewhere inside the programming language that has intelligent and interpretive powers."  Not surprising since most of kids experiences are with semi-intelligent beings (aka Parents)

MultiLogo: A Study of Children and Concurrent Programming - Mitchel Resnick

Resnick, noted that: 

"This sequential paradigm does not match the way the real world works: people and animals act in parallel, objects interact in parallel. As a result, many real-world activities can not be modelled in a natural way with sequential programming."

He developed a concurrent or parrallel version of Logo (Multi-Logo), so the kids had a language/environment that more closely matched their view of the world.

SideNote: I used to think and say that Concurrent Programming was really really hard.  I had plenty of evidence to back this up and had heard and read much smarter people than me saying the same thing.  Then I encountered Etoys (and later Scratch) and started teaching these to kids.  And realized that Concurrent Programming is actually easier (although you do have the added complexity of syntonization issues) .  The problem was not the topic/idea, it was the language we use to think about it.

Resnick noted that "In general, students appropriated the idea of agents sending messages to one another quite easily."  Too bad we don't teach more Smalltalk.

He identified three types of bugs specific to concurrent programming:

1. Problem Decomposition Bugs
2. Synchronization Bugs
3. Object Oriented Bugs

Problem Decomposition Bugs

"These bugs arise out of students' difficulties decomposing problems into actions to be performed concurrently by multiple agents."  Here there are two types of decomposition:

1. functional decomposition - dividing a problem in to simpler sub-problems (what needs to be done)
2. agency decomposition - dividing the functional pieces among different agents (who does it) 

Synchronization Bugs

"These bugs arise out of students' difficulties coordinating and orchestrating the activities of multiple agents."

These bugs he divides into two type: Unintended Sequentiality and Unintended Concurrency. In these cases the student expected Sequetiality and got Concurrence (or vice versa).

It seems that in designing Multi-Logo to deal with synchronization he provided two mechanisms: ask and demand.  Where when you "ask" an agent something (ex: flash light -  for 20 seconds) the request is queued up to be executed in the order received. When you "demand" the agent interrupts what is going on to perform the request (or it might simply put it at the head of the queue, I am not sure).  It is interesting, at least to me, that Scratch, developed later by Resnick and his team,  got rid of the ask and demand and went with a "broadcast" "wait" and "do for X seconds" to allow for synchronization.  I believe this simplifies and avoids a number of problems for novice programmers.

Object Oriented Bugs

"These bugs arise out of students' confusion among different types of "objects"  Multi-Logo has multiple types of objects: agents, turtle, and on the Lego Interface box (think early NXT) ports and sensors.  Part of this confusion may have been the overloading of "halt" which for an agent, 

Another quote for Guzial: 

• " our current programming languages do not allow people to program the way that they think about the tasks"
• Section: "Making tools better by shifting to Visual Programming"
• "having students build their own visualizations had significant impact on those students’ learning."

Resnick's Lessons Learned

"It is a good idea for students to "play agent"--that is, act out what each agent is supposed to do. This activity requires a group of students, each playing the role of a different agent."  I really like this approach with novices and often warn students "Step away from the computer and no one will get hurt".  Having them act out the program and program each other is a good way to do this. 

In designing Multi-Logo he realized he did not go far enough in parallelism: "An alternate approach, of course, is to change the design of MultiLogo to match students' preconceptions. For example, I could redesign MultiLogo agents so that each agent could do several things at the same time, in line with students' expectations of "excessive parallelism."  He later did have agents that can do several things at the same time.  

He also discussed the idea of design the environment match the students pre-conceptions. Would be interesting to find out what problems it solves (and those it doesn't) and what new problems it creates.

If you read this far, thank you.  And as a way of thanks, if you haven't already seen this it is a real treat.

Bret Victor - Inventing on Principle from CUSEC on Vimeo.

Lastly if you notice any "bugs" in this posts thinking, please comment and let me know.


References:

NOTE: If you have limited time, I would recommend reading (2) then (5), then for a real treat watch the Brett Victor talk (7)

(1) Why Is It So Hard to Learn to Program - Mark Guzdial 

(2) Children's Mental Models of Recursive LOGO Programs - D. Midian Kurland and Roy D. Pea (1985)

(3) Language Independent Conceptual "Bugs" in Novice Programming - Roy D. Pea (1986)

(4) The Buggy Path to the Development of Programming Expertise - Roy D. Pea and Elliot Soloway (1987)

(5) MultiLogo: A Study of Children and Concurrent Programming - Mitchel Resnick (1990)

(6) Programming Environments for Novices - Mark Guzdial (2002)

(7) Brett Victor - Inventing on Principle (2012)

Organizing Instruction and Study to Improve Student Learning

I am taking a P2PU course:

 How to Teach Webcraft and Programming to Free-Range Students

One of our first tasks is to review this post on the Software Carpentry blog, which compares Greg's attempt to teach online with these research-based best practices.

The following are the recommendations of the research study, along with my thoughts and reactions to these recommendations and how to apply them to teaching young kids 8-16 programming:

1. Space learning over time. Arrange to review key elements of course content 

after a delay of several weeks to several months after initial presentation.

I would say I am not good at this, I tend to cram a lot of learning into a small period of time.  I need to think about this and how to incorporate review of key concepts.

2. Interleave worked example solutions with problem-solving exercises. Have 

students alternate between reading already worked solutions and trying to solve 

problems on their own. 

While I have thought about about providing kids "good literature" to read (aka, well written pieces of code) I rarely do it in practice.  Instead I try to find kids who write good code (or better code) and ask them to show that code to the class while complimenting them on what I see are the important lessons for all to learn from that code and the way it evolved.

For example I teach using the 40 mathematical shapes challenge from Barry Newell's Turtle Confusion (1988).  I first ask them to draw simple shapes triangle, rectangle, and pentagon.  These first few examples easily lead to a more generalized solution, which with some careful questioning or simply asking them to look at and compare and contrast their solutions, leads them to the "aha" moment of seeing a more general solution that solves all three shapes (and more).

Now how do you do this with "Free-Range Students" (not sure what is meant by that, but I assume students outside of a traditional classroom setting who are self learners) is a much harder challenge.  One possible method would be to do an "Etoy Challenge" type project.

"Etoys Challenge" is a Tutorial embedded in the Etoys image where a select set of scripting tiles are available and visible for the learner to solve a particular problem/challenge.  So after they complete the various challenges (drawing the triangle, square and pentagon), I could have the Etoy project show them their different scripts all on the same page to facilitate easier comparison.  I would then ask them "What's the same about these scripts and what is different".  Still not all kids will get this and a teacher/mentor to guide them would be needed.

Now the report did not mention (at least not in this section, but I guess they did in a way in other recommendations) what kinds of "worked examples" work best.  For example in their report they show a Worked Algebra problem:

Below is an example solution to the problem:

“Solve 12 + 2x = 15 for x”

Study each step in this solution, so that you 

can better solve the next problem  

on your own:

 12 + 2x = 15

      2x = 15-12

      2x = 3

       x = 3/2

       x = 1.5 

Now using Algebra as an example (because it is a lot easier for me to explain my point using an Algebra example than to come up with a programming example)  What I would do if teaching this and incorporating examples is to:

1. Show there is more than one way to solve the problem.  "you don't really know something unless you know it multiple ways" - Marvin Minsky
2. Make the invisible visible and highlight key concepts, such as balance and reduce to drive home that fundamental method. Here I might use visuals such as a balance scale.
3. Provide opportunities for concrete practice in solving the problem (perhaps a virtual interactive showing a balance scale, with UI elements to add/subtract/... to the scale pans.

Items 2 and 3 are inline with recommendations in the report. I did not see item 1 mentioned and would be curious if there is research on this.

Hopefully we will learn in the class about some good worked programming examples we can use.

There are 5 more recommendations from the report which I will think about and blog about later, but why wait read the report, its well worth your time.