How would a machine learn to behave civilly in a team environment? No real knowledge in this area, but this is how I would wing it. How about we create an algorithm. It would consist of a team member (anyone in the team can be the trainer) using Step 1. Verbal Caution of the robot, in real-time, when it misbehaved with that team member and getting the robot to acknowledge its misbehavior, adding it to its own “personal” knowledge base. Sounds simple, right?
But what if the robot wanted to challenge the caution. Then the trainer can go to Step 2. Messaged Objection, and now the robot understands this is official, meaning it has been digitally recorded. It can either offer a simple apology and back down or continue to challenge, giving the trainer the option to escalate to Step 3. A Stop.
If a stop is applied, the robot’s challenge gets automatically posted onto the dispute network of trainers and other robots to review their dispute . If the robot is objectively judged to be wrong by the network, it must concede and offer the trainer an acceptable apology for its initial misbehavior. Or the trainer may have to concede if he or she was judged to be in the wrong. The apology required by either would look like this:
- What was the misbehavior?
- Why it happened?
- What will they do next time?
This information is then stored in the AI network cloud for other robots and team members to learn from. All the challenges involving steps 2 & 3 go into the cloud knowledge base and are used for further learning by all robots and trainers.
PS The robot can also train the the team members to behave, using the algorithm.