|
Bongard Problems sorted left have the keyword "spectrum" on the OEBP.
In a "spectrum" Bongard Problem, there is an evident way to assign each object a value (e.g. "size" or "number of holes"). Then, to determine whether an object fits left or right in the Bongard Problem, its value is compared with a fixed threshold value.
Spectra can be continuous or discrete.
A "spectrum" Bongard Problem is usually arbitrary, since there could be made many different versions of it with different choices of threshold value. However, sometimes a certain choice of threshold is particularly natural. For example, the threshold of 90 degrees in "acute vs. obtuse angles" does not come across as arbitrary. And in BP2, the spectrum of values ("size") is vague, so much that the fuzzy threshold, of about half the size of the bounding box, does not seem arbitrary.
A spectrum Bongard Problem may or may not have the following properties:
1) The values assigned to objects are precise.
2) The threshold value between the two sides is precise.
3) The threshold value is itself sorted on one of the two sides.
Each of the latter two typically only makes sense when the condition before it is true.
If a spectrum Bongard Problem obeys 1) and 2), then it will usually be precise.
For example:
"Angles less than 90° vs. angles greater than 90°" is "precise".
If a spectrum Bongard Problem obeys 1), 2), and 3), then it will usually be allsorted.
For example:
"Angles less than or equal to 90° vs. angles greater than 90°" is "allsorted".
Discrete spectra usually satisfy 1) but do not satisfy 2). In a discrete spectrum Bongard Problem, there isn't one unambiguous threshold value. Consider "2 or fewer holes vs. 3 or more holes". (Is the threshold 2? 3? 2.5?)
In an especially extreme kind of spectrum Bongard Problem, one side represents just a single value, just the threshold value. For example, "right angles vs. obtuse angles." In certain cases like this the threshold is an extreme value at the very boundary of the spectrum of possible values. For example, consider "no holes vs. one or more holes." Cases like this might not even be understood as two sides of a spectrum, but rather the absence versus presence of a property. (See the keyword notso.)
Even more extreme, in some Bongard Problems, each of the sides is a single value on a spectrum. For example, BP6 is "3 sides vs. 4 sides". We have not been labeling Bongard Problems like this with the keyword "spectrum".
After all, any Bongard Problem can be re-interpreted as a spectrum Bongard Problem, where the spectrum ranges from the extreme fitting left to the extreme of fitting right. | 
