I agree that Gould (as was common for him) was letting his political agenda undermine his reason and science in regard to the issue of IQ generally and IQ-genes in particular.
I agree his comment carries all the poor reasoning of Pascal's Wager and that he ignores two of the four cells.
I think an important issue is that the answer for each of those 4 cells is likely to depend on whether the question is "Are within group intelligence differences genetic?" versus "Are between racial group differences genetic?"
As I have argued with you before, these are completely different questions that likely have different objective answers. In addition, the consequences of belief are likely also much different.
Focusing on the former question about within group differences Gould is wrong that "we don't know". We know with more confidence than for many of his own claims about evolution that differences in IQ partly reflect biological and partly reflect environmental influences. We know that at the aggregate level, each type of factor is a big enough that for any specific person, their relative IQ standing could be almost entirely one or the other.
So then what are the consequences of personally believing this, scientifically supported position, to be true?
I would say there are little to no inherent negative consequences of believing that within group differences are a mix of genes and environment, because it leaves a chasm of room for any person to believe whatever they want about themselves or any other particular person. For anyone that tests poorly, they can believe that its all environment and that they can improve. It also has no clear policy implications, other than implying the wrongness of any policy that treats either low or high scorers as though they are a homogeneous group that would benefit from the same approach, since we know that people with the same score vary in how the different factors are responsible for their similar scores. OTOH, there are potential positives of believing this scientific position, because it frees us from either ideological and unscientific extreme. It allows us to recognize situation where there is rather compelling evidence that a person is struggling due to biologically grounded low IQ. Also, a bio-based low IQ does not in any way imply a policy of abandonment. We know that environment can impact the IQ of anyone. So, any and all low IQ people could be helped by education, but they best educational approach and most effective method will depend on the source of their current IQ level.
If we assume that all kids can be brought up to higher achievement levels by merely changing their environment, then we will fail to create instruction that is effective for many of them whose biological constraints require that they approach learning via an alternative strategy.
The issue is highly analogous to accepting the reality of differences in spatial ability. There is growing evidence that differences in general spatial ability has a causal impact on learning complex ideas, especially in science where the phenomena are often about relations among objects in space over time, from physics and chemistry to biology, neuroscience, and even history. Mostly the same folks that deny the relevance of IQ, try to deny the relevance of spatial abilities. One problem is the highly reliable difference between genders in spatial ability. The desire to deny any difference other than genitals (a position we often see here), leads to the denial of spatial ability as anything but a by-product of sexist socialization surrounding sports and science. Boys are encouraged in sports and science and thus develop spatial ability as by-product rather than ability being a causal factor in why they do better in science and thus select into it.
The "its all sexism" belief could do harm females. It presumes that all that is needed is more encouragement and females will perform equally well. IF that isn't true, then they will do worse. In contrast, if we acknowledge that some people have lower spatial ability and it impact science understanding, then we can design learning materials that convey the critical info in a manner that depends less upon their ability to mentally simulate dynamic relations among objects.