A large component of interest rates is a risk premium that depends on the likelihood of a borrower defaulting.
Credit scoring significantly outperforms any other methodology for assessing default risk, including credit matricies and especially human assessments (humans are shockingly bad at assessing default risk, rarely much better than a biased coin flip).
It's all well and good to say we should get rid of credit scores, but because they are so much more effective at assessing default risk than other methods, the consequence will be significantly higher rates of default, which means more people in financial hardship, and higher interest rates generally (though especially to low risk borrowers who will now be assessed as having closer-to-average default risk)
It's also not often appreciated, but credit scoring is also the single best technique we have to stop people borrowing beyond their means and entering into financial hardship and debt spirals. There are other techniques that exist to identify at-risk borrowers but these alone aren't as good as an approach which also incorporates credit scores.
As for explainability, unfortunately the best credit models are trained using AI techniques, which results in low explainability (since the risk signals are complex and multivariate). Older GLM approaches can be used, but aren't as good, so if we want high explainability, the trade off is worse performing credit models, and thus higher borrowing costs.
I think much of the distaste for credit scores comes from their use for purposes other than making loans.
I think it's reasonable for people to want a society where past financial difficulties (self-induced or otherwise) do not make it difficult for a person to get a job or rent an apartment. That's probably a reasonable preference to impose through legislation even if a credit score has predictive value for the legitimate interests of employers or landlords.
The challenge would be that there isn't just one "credit score". The scores published by bureau's exist more for marketing and as a stick to improve consumer behaviour.
A credit bureau's real product is the credit file, which contains your history of inquiries, defaults, collections, court judgements etc. In some countries it also includes granular payment history information.
From this file, many lenders compute their own internal credit scores. They do this because the credit scores published by the bureaus are the likelihood to default on any loan, however in practice consumers are often more likely to default on certain types of debt than others. Also, many lenders have additional data points that can be considered which the bureau's don't capture, such as the structure of the loan.
If there's anything questionable going on (e.g. using variables which act as proxies for factors prohibited by ECOA or FHA) it will be occuring in these proprietary lender-specific models, however the parameters used in these models also embed commercial sensitive information about the behaviour of their customer base, so I doubt many lenders will be keen to release them.
So why not ask the prospective customers for the right to fetch and use their data? If it gives them better rates then people have plenty of incentives to allow them to continue.
Credit scoring significantly outperforms any other methodology for assessing default risk, including credit matricies and especially human assessments (humans are shockingly bad at assessing default risk, rarely much better than a biased coin flip).
It's all well and good to say we should get rid of credit scores, but because they are so much more effective at assessing default risk than other methods, the consequence will be significantly higher rates of default, which means more people in financial hardship, and higher interest rates generally (though especially to low risk borrowers who will now be assessed as having closer-to-average default risk)
It's also not often appreciated, but credit scoring is also the single best technique we have to stop people borrowing beyond their means and entering into financial hardship and debt spirals. There are other techniques that exist to identify at-risk borrowers but these alone aren't as good as an approach which also incorporates credit scores.
As for explainability, unfortunately the best credit models are trained using AI techniques, which results in low explainability (since the risk signals are complex and multivariate). Older GLM approaches can be used, but aren't as good, so if we want high explainability, the trade off is worse performing credit models, and thus higher borrowing costs.