The experience with COSINE has clearly shown that SiFAs need to have history keeping capabilities in order to solve design problems efficiently.
Even very simple history keeping would improve the system drastically. For example, if an agent keeps a record of its proposals that led to a conflict, it can avoid those in the future. There are many other opportunities to improve the system with history keeping.
Since there is no agent hierarchy in the SiFA paradigm, no agent has a global view of the design process. Hence, all history keeping is at an agent level. Agents can remember their past interactions with other agents, but no agent can keep a global history of the whole design process. If an agent hierarchy where some agents control others is introduced in the future, then global history keeping becomes possible.
Keeping history is just a simple way for agents to learn as the design process goes on, and to improve the overall performance of the system with time. There are other opportunities for agents to learn during the design by using various learning techniques as discussed in the next section.