Viable System Model (VSM)

VSM stands for the Viable System Model and it introduces the concept of a viable organization and its ability to survive in a changing environment. VSM consists of a number of systems that correspond to the needed roles in an organization for it to be viable and self-producing. There are four underlying principles for VSM according to Beer[1]. These four principles and relevant comments on my behalf are presented below.

Principle 1 (P1) – Managerial, operational and environmental varieties, diffusing an institutional system, tend to equate; they should be designed to do so with minimal damage to people and to cost

P1 talks about the states of a system and suggests that any changes to the states/variety in the system need to be done with minimal disturbance to the entire subsystem. We can take a simple example being a light switch that has 2 varieties/states (off and on). In general, a control system must have the same level of variety in order to monitor and govern a system. When it comes to machine learning, especially in unsupervised ML, the system states can sometimes be so varied that the controlling system cannot comprehend as to what goes on inside the system. eg: FAIR[2] turned off their chatbot creation since it invented its own language[3]. In the future, P1 might need to be changed to incorporate self-correcting mechanisms to allow complex systems to survive.

Principle 2 (P2) – The four directional channels carrying information between the management unit, operations, and the environment must each have a higher capacity to transmit a given amount of information relevant to variety selection in a given time than the originating subsystem has to generate in that time.

Of course, P2 deals with capacity and capability management. Again, I foresee that instead of increasing the capacity for the directional channels to receive & transmit information, the local units should be better prepared to take course correctional decisions by itself without generating extra system overhead. In terms of projects, for instance, the monitoring and control mechanism itself should allow only relevant details to be transferred to other concerned organizations. For example, a portfolio manager is not really interested in the velocity of a team involved with the project and such information even if transferred is useless increasing complications to the already strained communications between a project manager and a portfolio manager.

Principle 3 (P3) – Wherever the information on a channel capable of distinguishing a given variety crosses a boundary, it undergoes transduction; the variety of the transducer must be at least equivalent to the variety of the channel.

P3 very much is in agreement with my statements in the previous paragraphs. The information generated goes through transduction and/or transmutation to satisfy the variety of the channel. Again, most of the information needs to be transmogrify with relevance to the channel and not just superficially wrapped.

Principle 4 (P4) – The operations of the first 3 principles must be cyclically maintained through time without hiatus or lags.

Again, from my perspective, if the system itself has been given the autonomy for course corrections, only the relevant information needs to transmogrify resulting in an overload load reduction as well as the extra capacity generation for tasks that involve collaboration and coordination. From a VPP (Viable Project Portfolio) perspective, it really means intervention when required and asked for. Most of the time, the specific point in time reporting aggravates and shows symptoms but not a cure.

What do you guys think? Shouldn’t these 4 principles be updated/adapted to the ideas mentioned in this post?

[1] – Stafford Beer (2008). Diagnosing the system for organizations. St. Gallen Malik Management Zentrum C.
[2] – Facebook’s Artificial Intelligence Research lab
[3] – ,