When Aiken devised the Mark I, why did he decided to separate data and instructions? It was not mentioned in Wikipedia (or in any other searches I've looked) on how or why Aiken separated data and instructions.
According to the Wikipedia article on the Harvard Mark I, the data memory consisted of storage for 72 23-digit numbers, and instructions were read from a paper tape loop.
That's the how.
The why is most likely that he hadn't seen the advantages of putting instructions into read/write memory. Also random access read/write memory was extremely expensive to implement ... compared with sequential and / or readonly memory.
Remember we are talking about a machine that was build from electro-magnetic switches and relays, and could do 3 additions per second.