Partial Associativity in Latin Squares
Latin squares arise from the multiplication tables of groups, but the converse is not true in general. Given a Latin square A, we can define a group operation giving A as its multiplication table only when A satisfies a suitable associativity constraint. This observation leads to a natural question concerning the '1%' version: if A is only partially associative, can we still obtain something resembling a group structure? I will talk about some joint work with Tim Gowers on this question.