I’m currently taking a course with Luca Dellanna covering:
- dynamics of antifragility
- the adaptive brain
- group dynamics
- designing for antifragility
The first session was about antifragility, and was extremely helpful in my own understanding and cleared up a few things that I had found difficult to convey to others.
The first thing that became clearer was how to visualise the states fragility and antifragility on a 1-dimensional scale. Dellanna uses a scale from fragile to antifragile. Unlike Taleb, he prefers not to include robust in this scale. I think this makes a lot of sense.
A recap about antifragility
Fragility doesn’t like variance. Fragility likes comfort but its precisely the comfort that causes fragilization over the long term. This implies always staying in the yellow zone (above).
Antifragility accepts (appreciates!) stressors / small damage so long as they are sparse – that is, not locaclised, and so long as there is growth resulting from those stressors. This causes the stregnthening / anstifragility over time. Green area.
Even the antifragile is fragile at a certain point, defined here as “functional impairment”, shown in red.
Within an adaptive system the points of fragility can vary at different levels of the system.
Dellanna gave a brilliant example of a business. Let’s say the business is suffering and a reaction to this stress is to get rid of staff. The business is still operating in it’s ‘green zone’, i.e. reacting, but where this zone starts is precisely where the red zone starts for the employee who ends up getting fired.
The diagrams are purely 1-dimensional to aid understanding. Of course any system will have multiple levels , mulitple dimensions and can be fragile in some areas and not in others. The diagram above can be broken down to more sub levels.
Increasing antifragility
To increase antifragility, we have three approaches:
- training (upping the exposure to stressors in the green area
- moving the threshold left i.e. the line between [no damage] and [spare damage] so that green area is larger) by,
- fractilizing the system (indepedent subsystems rather than monolithic)
- removing buffers (allowing problems to surface and dealing with them), or allowing harm to happen.
- moving the threshold right i.e. between [sparse damage] and [functional impairment] by,
- increasing redundency
- decreasing correlation
The difference between buffer and (extra) redundency
In software development, a buffer might be the time buffer between releasing new software every six months. By removing this buffer and releasing software every two weeks, it will surface all the impediments that need to be removed in order to develop software in a lean manner and continously improve.
On the other hand, the deliberate redundency might be hiring extra people (more than what is required) to either jump in for staff shortages or in case new opporunities have to be spun up.
Why is this all important? Because how systems reacts to an event is more important than the event itself.