There is no assumption since in this thought experiment the two groups have been defined to be how they are as an example which does not necessarily reflect reality 100%.

This means diversity of thought necessarily follows form the example given as it was defined as such (group 1 has diversity of thought, group 2 not).

The importance of diversity of thought can be illustrated by using machine learning; if your learning factor (diversity of thought) is too low, not all solutions are explored, instead a local minima is found and then reinforced by the ML algorithm. If the learning factor is higher then the algorithm can freely explore for other local minima and maybe even find the absolute minima. (If it's too high then you also don't get anything useful either but that's not quite the point)

Diversity of thought is important for a company that wants to make better product, if you only have people who think alike, regardless of their other properties, they will not find the best optimal solution, only a local optimal solution.