How to Price Distraction in Decision Making?

Distraction seems like a strange thing to price, but if you consider it in the context of decision making, it definitely has a price – if the decision maker is distracted, it will take them more effort to reach their goal. A simple simulation can be done to show just how much distraction may cost relative to a case when the agent’s attention is directed to the goal.

This text is part of the series on the design of decision governance. Decision Governance refers to values, principles, practices designed to improve the quality of decisions. Find all texts on decision governance here, including “What is Decision Governance?” here.

Let’s say we have an agent who moves through a grid from a starting position to a goal position. The agent does not know the goal position. 

We then put in place decision governance that at every step directs the agent’s attention to the goal position. If that governance is absent, the agent is distracted and is choosing randomly its next steps. In addition, and simply to avoid very large numbers of steps that the agent could take, there will in both cases be decision governance that ensures the agent avoids revisiting positions, unless there is no other choice.

The price of distraction is the difference between the number of steps the agent will take on average to reach the goal, in each of the two cases: when the agent’s attention is directed to the goal position, and when it isn’t.

Two sets of simulation runs are executed, each of 30 runs. The grid that the agent moves through has 30 x 30, or 900 positions. The starting position is in the upper left corner, the goal in the lower right corner of the grid.

The histogram below shows the case when the agent is distracted. On average, it took about 2100 steps to reach the goal, with a standard deviation of about 1660 steps.

The following shows the paths taken when the agent is distracted. What you cannot see below is that the agent revisited some of the positions.

If the agent is not distracted, that is, its attention is directed through decision governance to the goal, the histogram over 30 simulation runs is quite different. It is shown below. The agent took on average about 100 steps with a standard deviation of about 25 steps. 21 times fewer steps on average than when distracted in the way I defined distraction above.

The runs expectedly show simpler paths.

Distraction can be worse – we can distract the agent not by having it randomly choose the next step, but by directing its attention away from the goal. The impact is simple to anticipate – the number of steps will increase further.
Using the decision governance design space terminology from the text here, the parameters governed in the simulations above are Attention and Memory, across all stages of the decision process. This is shown in the image below.

Having more runs in the simulations does not change the observations above. The histogram below is for 1,000 runs of the distracted agent.

The following histogram shows steps across 1,000 runs for the agent whose attention is directed to the goal at each step.