In a self-driving car, inputs from the surrounding environment are processed, and outputs of speed, steering and navigation of the vehicle are modeled and executed. The assessments of risk, the prediction of outcomes and the ensuing actions happen with varying levels of automation depending on the level of automation.
The car model
- Level 0 – no automation; the driver drives.
- Level 1 – hands on; there is shared control of speed, steering and braking.
- Level 2 – hands off; driver is still alert and ready to intervene immediately.
- Level 3 – eyes off; the automated system is in control but can alert the driver to intervene in certain circumstances.
- Level 4 – mind off; the driver can go to sleep but only in certain areas.
- Level 5 – steering wheel optional; full control is with the system and there is no driver intervention.
The phrase data-driven decision-making is common. Yet, in the age of self-driving cars it is not a binary choice between whether a human or a system is driving the decisions.
The decision model
- Level 0 – no automation. The decision-maker drives. They will use inputs of knowledge, emotions and incentives, plus their instincts and experiences alongside any objective data they have available (which they are also free to ignore).
- Level 1 – hands on. Where there is strong trust in the information then aspects of decision-making can be strongly informed by insights and associated recommendations.
- Level 2 – hands off. Decision-making can be automated from the data but decision-makers still check the information, insights and recommendations. Decision-makers confirm actions.
- Level 3 – eyes off. Decision-maker relies on automated systems for decisions in a particular area, with the exception of times when an alert triggers them into corrective action.
- Level 4 – mind off. Fully automated prediction and decisions running in a system – e.g. automatic marketing journeys where messages and offers are automated based on existing customer preferences and purchases. This will still be limited to certain activities and these parameters can change.
- Level 5 – steering wheel optional. Full control is with the system and there is no decision-maker intervention. Complete trust in the predictive outputs of the automation and full autonomy with those building the system.
How well do you think the self-driving car metaphor stacks up as a way of looking at this spectrum of control between decision-maker driven and data-driven actions?