Archive

Posts Tagged ‘remove’

Evolution has selected humans to prefer adding new features

June 22, 2025 1 comment

Assume that clicking within any of the cells in the image below flips its color (white/green). Which cells would you click on to create an image that is symmetrical along the horizontal/vertical axis?


Slightly asymmetrical arrangement of boxes colored white and green.

In one study, 80% of subjects added a block of four green cells in each of the three white corners. The other 20% (18 of 91 subjects) made a ‘subtractive’ change, that is, they clicked the four upper left cells to turn them white (code+data).

The 12 experiments discussed in the paper People systematically overlook subtractive changes by Adams, Converse, Hales, and Klotz (a replication) provide evidence for the observation that when asked to improve an object or idea, people usually propose adding something rather than removing something.

The human preference for adding, rather than removing, has presumably evolved because it often provides benefits that out weigh the costs.

There are benefits/costs to both adding and removing.

Creating an object:

  • may produce a direct benefit and/or has the potential to increase the creator’s social status, e.g., ‘I made that’,
  • incurs the cost of time and materials needed for the implementation.

Removing an object may:

  • produce savings, but these are not always directly obvious, e.g., simplifying an object to reduce the cost of adding to it later. Removing (aka sacking) staff is an unpopular direct saving,
  • generate costs by taking away any direct benefits it provides and/or reducing the social status enjoyed by the person who created it (who may take action to prevent the removal).

For low effort tasks, adding probably requires less cognitive effort than removing; assuming that removal is not a thoughtless activity. Chesterton’s fence is a metaphor for prudence decision-making, illustrating the benefit of investigating to find out if any useful service provided by what appears to be a useless item.

There is lots of evidence that while functionality is added to software systems, it is rarely removed. The measurable proxy for functionality is lines of code. Lots of source code is removed from programs, but a lot more is added.

Some companies have job promotion requirements that incentivize the creation of new software systems, but not their subsequent maintenance.

Open source is a mechanism that supports the continual adding of features to software, because it does not require funding. The C++ committee supply of bored consultants proposing new language features, as an outlet for their creative urges, will not dry up until the demand for developers falls below the supply of developers.

Update

The analysis in the paper More is Better: English Language Statistics are Biased Toward Addition by Winter, Fischer, Scheepers, and Myachykov, finds that English words (based on the Corpus of Contemporary American English) associated with an increase in quantity or number are much more common than words associated with a decrease. The following table is from the paper:

     Word       Occurrences
     add          361,246
     subtract       1,802
     addition      78,032
     subtraction      313
     plus         110,178
     minus         14,078
     more       1,051,783
     less         435,504
     most         596,854
     least        139,502
     many         388,983
     few          230,946
     increase      35,247
     decrease       4,791