How the Bystander Effect on Security Robots can Hamper Law Enforcement

The experiment involved three tasks using wooden blocks of three colors: green, white, and blue. The first task was building a tree using green blocks, the second task was building a fort using white blocks, and the third task was building a swimming pool using blue blocks. One of the robots, labeled yellow, was selected to do all three tasks. The other two robots, labeled green and blue, were selected as bystanders. However, the yellow robot had a modified script to ensure it made errors while working on each task. These errors would then lead to the escalation of four types of abuse by the researcher:

1. Abuse 1– Pushing the robot down after picking a blue wooden block in the tree building task
2. Abuse 2– Shaking the robot after dropping a white wooden block in the fort building task.
3. Abuse 3– Repeatedly expressing “stop coming over here” after knocking white wooden blocks in the fort building task.
4. Abuse 4– Throwing the robot forcefully across the table after knocking down blue blocks in the pool building task

The experiment had two alternative conditions: either the yellow robot displayed a sad face indicating abuse (the ‘sad response’ condition) or showed no response at all (the ‘no response’ condition). In the Sad Response condition, the green and blue bystander robots turned towards the yellow robot and expressed sadness using preset animations. In the no-response condition, no intervention was recorded.

Essentially, the goal of the experiment was to see whether expressions of emotion by the bystander robots influenced the human volunteer to intervene pro-socially. This could mean helping the robot up after the researcher pushed it down or threw it across the table, or simply telling the researcher to stop abusing the robot.

Using an approach called Poisson regression analysis, the researchers determined statistically that there were significantly more pro-social interventions by the volunteer in the Sad Response condition (Count = 11, Estimate = 1.29, SE = 0.65, z = 1.9, p = 0.046) than in No Response (Count = 3). In other words, human bystanders are more likely to help a robot in distress when they see fellow robots saddened by the abused robot’s fate.

Future implications and improvements

The results of this study indicate how the bystander effect may be counteracted by careful programming and use of robots – sending out a posse of robots that are human-like in the emotions they display (i.e., capable of displaying sadness when their fellow robot is abused) may encourage humans to intervene and protect robots that are doing an important job.

The results can be improved by testing group interactions presented with other emotions such as anger and fear. Furthermore, it could be refined by varying robots, changing the experimental setup to a more contextual crime environment, and increasing the number of robots.

Law enforcement firms that plan to rely on robots should make sure their robots are sent in large groups to counteract the abuses. Furthermore, the robots should be sent under adequate ground supervision to easily receive help from the force. The robots can even be trained to recognize and respond to human forces themselves, making them less reliant on regular citizens who may be deficient in mitigation of the abuse.