Soft goals
Unlike behavioral goals, soft goals prescribe preference among alternative systems behaviors. They are more fulfilled along some alternatives less along others.
Considers the following goals in our meeting scheduling systems, for examples:
Interactions with invited participants should be limited as much as possible.
This goals prescribes that behaviors where there are fever interaction (for example though e-agenda access ) are to be preferred to behaviors where there are more interactions (for example through e-mail requests and reminders). Here is an additional sample of soft goals for our running case studies:
The stress underpinning the working conditions of train drivers should be reduced.
Passengers should be better informed of flight connections and airport facilities.
The meeting scheduler software should be easy to use by administrative staff.
The workload of library staff members should be reduced.
A soft goal cannot be established in a clear-cut sense. For example, we cannot say in a strict sense whether such or such a system behavior in isolation satisfies the stress-reduction goals or not. We might, however, say that one system behavior may reduce stress further than another. Put in more general terms , the phrase ‘goal satisfaction’ does not make too much sense for soft goals, as we cannot observe that they satisfied by some behaviors and not satisfied by others. The phrase goal satisficing is sometimes used instead; a soft goal is more satisfied in one alternative than in another (Chung et al., 2000).
Soft goals are therefore used as criteria for selecting one system option among multiple alternatives -in particular, one goal refinement among alternative ones, one risk countermeasure among alternative ones, one conflict resolution among alternative ones, or one goal responsibility assignment among alternative candidate agents. We will come back to this in Chapter 16
In a way similar to Achieve and Maintain behavioral goals, we may prefix a soft goal name by a keyword indicating a corresponding pattern, such as:
Improved [Target conditional],
Increase [Target quantity], Reduce [Target quantity],
Maximize [objective function], minimize [objective function].
As already introduced in Section 4.2.1, the specification of a soft goal should in addition clearly state a fit criteria to ensure its measurability.
Soft goals are not the same as goals whose satisfaction is required in X% of cases. The latter behavioral goals sometimes called probabilistic goals. Unlike soft goals, we can observe whether a probabilistic goals satisfied or not ; satisfaction is only required ‘ on some average’ due to limited resources, possibility of uncontrollable failure and so on. For example :
The meeting scheduler should determine schedules that fit the diary constraints of all invited participants in at least 80% of cases.
The distance between two consecutive trains should always be greater than the worst-case stopping distance of the train in front, with a mean time between hazards of the order of 10' hours.
Thus, the phrases ‘satisficing’ and ‘partial degree of satisfaction’ do not mean the same. The current state of the art in evaluating, validating and verifying probabilistic goals is still fairly limited, although such goals are frequently found in software projects (Lettier & van van Lamsweerde, 2004).