Once the agent notices that what is already stored in its target entity is not identical to what it wants to put in that entity, and indicates a conflict, the agent has to decide if what it wants to store is compatible with what is already there. If they are compatible then there is no conflict although they may not be identical. It is not sufficient to base conflict detection simply on checking if values, estimates, evaluations, criticisms, or praises are exactly identical.
This situation is best exemplified in the case of values. Suppose a design parameter can take on continuous numeric values, and its current value has been set to 5.00 by selector A. If later selector B calculates a value for the same parameter from a different point of view and comes up with 5.01, it indicates a possible conflict, but it has to do more work in order to decide if there is indeed a conflict or not. Selector B needs to know how sensitive the value of the parameter is. This sensitivity also depends of selector B's point of view, or what selector B is trying to optimize. If allowing up to 1% changes in the value of the parameter is acceptable by selector B, then there is no conflict. Selector B can just leave the value as 5.00 and does not need to insist on 5.01. Similar situations can occur with other agent types and entities as well.
For agents to be able to detect incompatibility conflicts, they should have the ability to decide if something that is stored in their target is compatible with what they produce. Target entities may be numeric or symbolic. Value entities would be numeric for parameters like length, but symbolic for parameters like color or material. Estimates would generally be of the same type as the value of the parameter. Evaluations may be symbolic such as good, average, bad, or numeric if the quality is measured as a percentage. Critiques and praises may contain both numeric and symbolic information.
If the target entity is numeric, then the agent should have either a percentage or a fixed value by which the numeric value, estimate or evaluation it produces may be relaxed. For example, the agents wants the value to be 50, but can relax 5%, so anything between 45 and 55 is fine, or relaxation by only 3 units is possible so anything between 47 and 53 is acceptable.
If the target entity is symbolic, there are two possibilities. If there is an ordering to the symbols, such as bad, average, good, then the agents can have a constraint such as ``anything better than average is acceptable''. If there is no ordering to the symbols than the agents need to have a list of things that they find acceptable.