Survey Explores Human Understanding of Computer Arithmetic

Programmers of any kind are encouraged to take the survey at http://presciencelab.org/float

Although it’s tempting to think that computers can do anything, they are not without their limitations.

Peter DindaFor example, computers’ limited memory makes them unable to compute real numbers. So, when a computer program attempts continuous mathematics, it must use floating-point arithmetic, which looks to the programmers like arithmetic with real numbers but is actually an approximation.

“A computer has a fixed amount of memory and that has consequences for the representation of numbers,” said Peter Dinda, professor of computer science in Northwestern’s McCormick School of Engineering. “Computers fundamentally cannot do continuous mathematics, and, indeed, the numbers commonly used in programs, including those in science and engineering, to approximate real numbers are only a few bytes long.”

Dinda and his team are examining how software developers understand floating-point arithmetic and compensate for its limitations. So far, Dinda’s team has rigorously surveyed users from across Northwestern, other universities, and national laboratories. Now it is opening up this survey to a broader audience.

If you have experience in computer programming in any role (including, but not limited to software development/engineering, engineering, scientific research, finance, management, etc.) and would like to help Dinda with his floating-point arithmetic research, please visit http://presciencelab.org/float to complete the survey.