Communications of the ACM

practice

SOFTWARE DEVELOPERS SPEND 35%–50% of their time validating and debugging software. 1 The cost of debugging, testing, and verification is estimated to account for 50%–75% of the total budget of software development projects, amounting to more than

$100 billion annually. 11 While tools, languages, and environments have reduced the time spent on individual debugging tasks, they have not significantly reduced the total time spent debugging, nor the cost of doing so. Therefore, a hyperfocus on elimination of bugs during development is counterproductive; programmers should instead embrace debugging as an exercise in problem solving.

It is more appropriate for programmers to focus their
efforts on acquiring and encouraging effective learning
strategies that reduce the time spent
debugging, as well as changing the way
they perceive the challenge. This ar-
ticle describes how effective problem-
solving skills can be learned, taught,
and mentored through applying re-
search on the psychology of problem
solving, conducted by Stanford’s Carol
Dweck and others.
In the 1974 classic, The Elements
of Programming Style, Kernighan and
Plauger wrote, “Everyone knows that
debugging is twice as hard as writing a
program in the first place. So if you are
as clever as you can be when you write
it, how will you ever debug it?” 6 Is de-
bugging really twice as hard as writing
a program? If so, why should it be?
To answer these questions, let’s first
consider why bugs occur. In a 2005
paper, Andrew J. Ko and Brad A. Mey-
ers suggest that bugs occur as a result
of “chains of cognitive breakdowns...
formed over the course of program-
ming activity.” 7
At a fundamental level, all software
describes changes in the state of a sys-
tem over time. Because the number
states and state transitions in soft-
ware can have combinatorial complex-
ity, programmers necessarily rely on
approximations of system behavior
(called mental models) during develop-
ment. The intent of a mental model is
to allow programmers to reason accu-
rately about the behavior of a system.

Since mental models are approximations, they are sometimes incorrect, leading to aberrant behavior in software when it is developed on top of faulty assumptions. Neophyte programmers experience this problem to a greater degree: their models of the language and development environment are necessarily incomplete, and even a simple syntax error can be a major blocker for someone learning to program. This is not limited to novice programmers. The C11 language specification is just north of 700 pages, for example. How many systems programmers understand the language completely?