Carrot, Stick, and Battery


Every change to a user’s (developer’s) behaviour is encouraged by some amount of carrot and some amount of stick.

There is always a switching cost for habits and workflows. In the case of carrot-motivated changes, the developer perceives enough value in the change that they make the change themselves with no additional encouragement.

When a user doesn’t perceive enough benefit in a change that they willingly make it themselves, the team has to implement a “stick” — extra incentive to make the change, typically through negative reinforcement of the old behaviour.

The additional element here is the fusion of “Carrot and Stick” with Tobi’s concept of Trust Batteries. Consider that each “Stick-based” change drains our team’s Battery, and each carrot-based change charges it. Resistance to stick-based changes increases as the battery drains, as does skepticism of carrot-based change.

Thus, it’s desirable to maintain a healthy weight toward carrot-based change and spend our battery on stick-based change only when truly necessary.

To elaborate this idea even further, for a proposed adaptation , take to be perceived benefit of the adaptation and to be the perceived adaptation cost. Then is the perceived value of an adaptation in:

A developer will only adopt if , in which case we can call a “carrot-based” adaptation. If , we have to apply an extra incentive, in the form of negative reinforcement of the old behaviour ( ) in order to make the effective value positive — making this a “stick-based” adaptation:

We define a developer’s trust battery to be a sum of all historical value minus “stick , subject to two different decay factors and , where (slower decay) to account for Negativity Bias.

In observing that developers are likely to be skeptical of changes from a team that they have a depleted trust battery with, we note that the perceived benefit can be decomposed. Let be the portion of perceived benefit not biased by the trust battery, and let be the value of the trust battery, with as some constant factor:

Putting this all together, we get:

The most interesting thing to note here is that there’s a feedback loop in subsequent values of : An initially high value of makes it easier to apply adaptations without using stick ( ), which makes it easier to continue to increase . Conversely, a low trust battery tends to require stick, causing continued erosion of trust. Additionally, because of the exponential decay factors, old reputation can be quickly overcome by a series of more recent failures or successes, but failures are harder to wipe away than successes.