Can Robots Ever Be Justified in Lying?

Robots and lies are becoming increasingly intertwined as AI and robotics advance rapidly. While policymakers and regulators may find it difficult to keep up with these developments, scientists are trying to understand their implications. A new study, published by Frontiers in Robotics and AI, investigates how people feel about the potential for robots to lie and deceive their users. The study presented scenarios where robots intentionally lied to people, focusing on two main goals. The first was to determine if certain lies are acceptable. The second was to explore how people might justify these lies if they found them acceptable. The investigation was driven by a larger debate surrounding AI and robotics, specifically the rights and responsibilities these technologies might have if they were to become sentient. If it’s acceptable for humans to lie to protect someone from harm, could the same apply to robots? According to the study, in some cases, the answer could be yes. There are three types of lies robots might use to deceive people The study identified three primary ways in which robots are likely to deceive people. The first method involves a robot lying about matters unrelated to itself. The second method involves a robot concealing its own capabilities, while the third involves a robot falsely claiming abilities it does not possess. To investigate these potential deceptions, researchers created scenarios featuring these types of lies and presented them to 298 participants through an online survey. Participants were asked to evaluate whether they considered the robot’s behavior deceptive and if such behavior was acceptable. Additionally, they were questioned about whether they thought the robot’s actions could be justified. To respondents, Type 1 lies are justified whereas Type 2 and Type 3 are not It’s important to note that all hypothetical scenarios presented in the study were considered deceptive by respondents. However, people generally approved of type 1 lies, which involve robots lying about something other than themselves, but not type 2 or type 3 lies. For context, 58 percent of those surveyed deemed type 1 lies acceptable if they prevented harm to someone. One scenario involved a robot lying to an elderly woman with Alzheimer’s, telling her that her husband was still alive. Respondents justified this by stating that the robot spared the woman from a painful memory. In contrast, respondents disapproved of type 2 and type 3 lies, which involved robots claiming abilities or emotions, such as feeling pain, that they did not have. Interestingly, when it came to these disapproved lies, blame was placed on the robot owners and programmers rather than the robots themselves.