A soft variable in system dynamics is a dynamic variable that is difficult to quantify. Although the concept is known to change over time, it is unclear how it can be assigned a numerical value or what that value would mean. A soft variable contrasts with a hard variable whose values are clearly defined and measurable.

By their very nature, soft variables, or soft concepts, are difficult or even impossible to measure. Yet, their inclusion in a model is often a matter of necessity, as they are known to be a part of a chain of cause and effect.

Example of a Soft Variable

Think of a public riot. The size of that riot can encourage feelings of enthusiasm on the part of the rioters, encouraging them to recruit more to the cause. The size of the riot is easy to enumerate – the number of rioters at a given time. It is a typical hard variable. However, the enthusiasm of the rioters is much harder to quantify. What are the units of enthusiasm? Rioter enthusiasm is a candidate for a soft variable.

The temptation is to leave candidates for soft variables out of a model, as calibration would be difficult.
However, the much-quoted comments of Jay Forrester (1961, p. 57) should encourage anyone to resist such a temptation: “To omit such variables is equivalent to saying they have zero effect – probably the only value that is known to be wrong!”

Soft Variable as a Stock

The above example on riots is discussed in Hayward et al. (2014). The main dynamic hypothesis was that new rioters were recruited through existing rioters according to the current enthusiasm of the rioters. In addition, enthusiasm is encouraged by the numbers joining the riot. These hypotheses can be turned into a stock-flow submodel, figure 1. R1 – the more rioters, the more join the riot. The rioters have personal influence over non-rioters. R2 – the more rioters, the more enthusiasm, the more effective they are in recruitment.

Figure 1: Recruitment to a riot through rioters and their enthusiasm.

Rioter enthusiasm is a candidate for a soft variable as it is unclear how it is measured. It is also a candidate for a stock as enthusiasm has persistence. Even if people leave the riot, the rioters do not lose enthusiasm straightaway. Although enthusiasm is generated by the number of rioters, it declines over time naturally in the absence of generation. Thus, the item, rioter enthusiasm, can be represented by a stock, figure 2:

Figure 2: Rioter enthusiasm as a stock

A fuller discussion by Hayward et al. (2014) and on Identifying a Soft Variable.

Measurement of a Soft Variable

By its very nature, a soft variable is difficult to quantify. It is sometimes called an “unquantified variable”. Nevertheless, its inclusion in a stock-flow model, such as that in figure 2, assumes it can be assigned numerical values. To help with quantification we could ask some questions

  • What does a value of zero mean? In the rioter model, it is easy to see that zero enthusiasm means the rioters have no enthusiasm for recruitment, which effectively switches off loop R2.
  • Could the value of the soft variable increase indefinitely? If not, then it can be assigned a maximum value. In the rioter model, figure 2, there is a good case for assuming there is a maximum to enthusiasm; see Hayward et al. (2014).

For further information, see Constructing a Soft Variable.

Use of a Soft Variable

Identifying and measuring a soft variable can be tackled in isolation from the rest of the model. However, as figure 2 shows, the variable needs to be connected to other model elements. Two questions need to be asked.

  1. How does the number of rioters generate enthusiasm? Is it linear, or are there non-linear mechanisms involved?
  2. How does rioter enthusiasm influence recruitment?

Extra converters need to be included to model the connections between the soft variable and the rest of the model, figure 3. These elements are sometimes called transmission functions. They “transmit” information between the other variables and the soft variable, handling scaling and units

Figure 3: Rioter enthusiasm with transmission functions

See see Hayward et al. (2014). More information about soft variables can be found on the pages:

  • Identifying Soft Variables
  • Constructing Soft Variables
  • Using Soft Variables

References

Forrester, J. W., (1961). Industrial Dynamics. Pegasus Communications, MA: Waltham.

Hayward J., Jeffs R.A., Howells L. & Evans K.S. (2014). Model Building with Soft Variables: A Case Study on Riots. Presented at the 32nd International Conference of the System Dynamics Society, Delft, Netherlands, July 2014.