@Anonymous wrote:There is something that is lost here...
Human brains can compute data not only faster than a computer can, but more efficiently. When you have a team of data scientists and software engineers together... well, that's much more computational power than you can imagine.
When we write software, it is imperfect. There are things we can do really well now, even with algorithms, but still not as well as the brain. Information in a computer currently only exists in bytes, in which that byte can be one of two values - 1 or 0. The more substantial the data, the greater quantity of 1's and 0's there are to process. With increased complexity comes increased time, and believe it or not, there are some processes that require such a substantial amount of data churning that it can exceed what our computers are even capable of processing.
We know that newer algorithms will be even more robust than what we've got now, but they still won't be perfect.
A set of numbers that are relatively similar, versus a complex web of data points and structures... it's just an entirely different ballgame.
Your general starting point is correct. The issue with the computers is, a programmer has to set up the inputs that the algorithm will consider, and to develop the program which is going to make some sort of calculation from those inputs. The more inputs that are available, and the more nuances that the programmer tries to include, then the more likely the output is unpredictable.
Humans only seem to calculate faster. What humans are actually doing is pulling in other factors, other assumptions, other biases into their "decision process". Humans are able to read a credit application, but also to look at the person across the desk, and include that bias into their decision. A computer can only deal with the data fields that are available to it.
Managers at my work are constantly frustrated because they have a business process idea that they would like the computers to handle. Humans can pick those up and adapt because they can think around the roadblocks, the issues, the communication blocks. Humans, however, cannot handle the volume of data that many of these processes generate. However, until you can absolutely nail down every possible variation you want to deal with, and tell the computer exactly what to do in every variation of the situation [and this means you MUST limit the variability of input and MUST limit the expected capabilities] the computer will stop dead and not know what to do.
This is relevant to the thread, because OP is bringing in judgemental explanations of particular situations. But the computer does not have access to those judgements that are being used to explain the differences. A pattern of "quick lates" could be just the start of "years of lates". Who knows how to separate the two until years have passed? And by that time, either there have been no additional lates, or there are "surprise" lates that in hindsight were indicated by the early bad patch.
My thinking...if what you mention is a 10-15 point spread and both parties (perfect and spotty) will eventually have access to the same benefits of great credit such as better interest rates and better trade lines, what does it really matter? What benefit does the perfect person gain by suppressing the spotty person who drastically changed their behavior? Bragging rights of a slightly better score?
I understand there are inequities in the scoring system, but in the end what is really gained if both parties have access to the same benefits? Also, what is lost to the better profile by allowing another to achieve a better score for themselves? Seems punitive to me.
Am I missing a something?
@Gidgetmom wrote:My thinking...if what you mention is a 10-15 point spread and both parties (perfect and spotty) will eventually have access to the same benefits of great credit such as better interest rates and better trade lines, what does it really matter? What benefit does the perfect person gain by suppressing the spotty person who drastically changed their behavior? Bragging rights of a slightly better score?
I understand there are inequities in the scoring system, but in the end what is really gained if both parties have access to the same benefits? Also, what is lost to the better profile by allowing another to achieve a better score for themselves? Seems punitive to me.
Am I missing a something?
FICO scores are intended to be a single number ( within each FICO version since the criteria are different ) which is intended to simplify the bell curve measurement of risk of borrowers. So I don't think you are missing anything. These scoring models have to somehow make sense of millions of combinations of personal credit history data, but at the end of the day, each set of data in a credit file is only as good as the human being, the borrower, who generated those credit events. FICO is most directly useful in creating new credit card accounts, because the automation simply has to be efficient. For larger loans, it is only one step of many as the lender wants to better understand the person they are lending to. The refinements of 10-15 points would get washed out, in mortgage and auto loans, by the larger credit history that is directly reviewed by an underwriter.
So while the idea of borrower intent and long history of patterns is an interesting idea to try to include in a FICO model, that actually gets taken into consideration with manual underwriting, where the lender really cares about the amounts being lent out.