I never unquestioningly accept the results produced by machines and as much as possible try to find independent ways to check if they make sense. The following story may explain why.

When I was in graduate school, my doctoral thesis involved a lot of detailed calculations that required using a computer. This was in the days prior to the personal computer and we used massive mainframes, entering the programs and data using punch cards and later advancing to remote terminals. Because the computer programs I had written were so complicated and there were so many opportunities for making errors, as much as possible I would check its output in special, simplified cases where I could also do the calculations using just a pocket calculator.

There was one occasion where I simply could not get the two results to agree. After days and days of work trying to find the source of disagreement, going to the extent of doing elaborate calculations without even the calculator, I found the source of the problem. It turned out that my hand calculator had this bug that if you had a number in the display that had the digit 8 in the fourth decimal place, and stored this number in the memory, when you recalled this number, the 8 would have been replaced with a zero. It was a very specialized error, occurring only with the digit 8 and only in the fourth decimal place. Everything else was fine. When I told my thesis advisor what had caused the problem he was shocked and said, “If you can’t trust your own calculator, what can you trust?”

It was the kind of bug that could escape detection for a long time because the chances of it making a noticeable difference in a calculation was extremely small but it shook me up so much that after more than three decades I still remember the details of that story.

I was reminded of this when I came across this item about a ‘Wrongulator‘, a gag calculator that always gives you the wrong answer.

I am not sure how it works. I would think that a calculator that is invariably wrong would be easy to detect unless you are totally innumerate. It also depends on how wrong it is. To fool someone, the error would have to be subtle, like my own experience. If the wrongulator said that 4×6=543, that would be easily detectable, whereas one that returned the answer of 26 may fool some.

I actually don’t like gag gifts like this. They could have very serious negative consequences in the hands of innumerate people who accept unquestioningly whatever machines tell them.