Comment by finite_jest | Hacker News Reader

Comment by finite_jest | Hacker News Reader

Parent Full thread

Interesting observation. I have not read the whole document, but based on Table 6.2 (page 34 of the PDF) it looks like almost all of the difference between ethnic groups comes from the people who took the test in French:

           |  % Successful (EN) |  % Successful (FR)
     White |        86%         |        84% 
 Non-White |        82%         |        55%

This seems quite strange to me. I am curious to know whether the French version is merely a translation of the English version, or there are other differences?

-------------------

PS. I just realized these so I'm gonna add them here.

There is a little bit of ambiguity about Table 6.2. Does the it French column refer to people who wrote the French version of the test (as the paragraph [33] of the ruling suggests), or does it refer to the languages the candidate spoke (as the existence of a "English+French" column suggests)?

Also, the vast majority of the candidates are Anglophones, so the overall difference is not as large as the above numbers may suggest. Overall, it's more like 80% (Non-White) vs 86% (White). (https://www.wolframalpha.com/input/?i=%28526%2B33%29%2F%2852...) Table 7 of the ruling has a more detailed ethnic breakdown.

I'm curious why a poor translation would more strongly affect non-white francophones? Why are the whites minimally/not affected?

> I'm curious why a poor translation would more strongly affect non-white francophones?

Canadian French, without assuming translation issues, might conceivably more strongly negatively effect non-White francophones, who may be more likely to be part of immigrant communities from (and natively speak the French of) various non-Canadian nations where French is commonly spoken, like (e.g.) Haiti.

Probably demographics; the white Francophones also speak English.

Does "non-white French speaker" actually mean indigenous peoples in this case? Is the actual problem being obscured by confusing labeling?

My guess was that it means people from African or Caribbean countries with a pre-existing relationship with the UK but whose first language is an uncommon one and whose second language is French.

Just an idea: maybe non-white francophones are immigrants from countries with a different dialect of French, so even if they are technically francophones they have a serious gap in understanding québécois. As French has an extra challenge in math by the way numbers are composed (ex: 80 is literally "4 of 20", not "8 tens") that makes decimal calculations a bit harder than in English, it can skew the results to the observed numbers.

Non-white French speaking populations are probably very distinct. Maybe immigrated from French-speaking former colonies? Though if every teacher in a French district takes the French test, then it would include any internal immigration in Canada...

From the practice test, some are weird Qs in english, let alone french.

> What is 19 thousands + 7 tens?

Idk French, but I wonder if white teachers are likely to be more familiar with english-to-french translation oddities. An immigrant might be baffled by what a Quebecois is simply annoyed by.

Reply on news.ycombinator.com