AI as a negotiator?
Today, we are beginning to learn negotiation from AI, but tomorrow we may have to trust them to negotiate on our behalf.
Happy Brief Thursday!
In this week’s Brief, I will share some latest developments in AI and negotiation.
Featured: AI as a negotiator?
Have you ever played the video game Detroit: Become Human? It’s a video game where you play as the androids in the world set in 2038, where robots are starting to become humans, if not more advanced. At the beginning of the game, there’s a hostage crisis, and the negotiation expert sent in to deal with it was an android! The mother of the hostage freaked out when she realized that an android was there instead of a human negotiator. This scenario raises some interesting questions about the use of AI in high-stakes situations.
Now, you probably think: that’s just a video game. However, AI is already better at negotiation than humans in some cases. For example, Meta, the parent company of Facebook, developed an AI named Cicero that can play the first world war strategy game Diplomacy. The AI is so good that it can participate in game conversations and negotiations with human players without them realizing that they are talking to a machine. Cicero learned to play Diplomacy through trial and error and by mimicking human players, and it has been further enhanced with a language component that enables it to interpret and generate messages. The AI has entered anonymously into 40 games of Diplomacy hosted by an online league of human players and sent 5,277 messages over 72 hours of gameplay. Almost nobody realized they were communicating with an AI. Cicero’s success shows the potential for AI to excel in tasks that require negotiation and strategic reasoning. [References: 1, 2]
Another piece of evidence supports that AI may become a better negotiator when it can infer what people think. Stanford University’s computational psychologist, Michal Kosinski, has used standard psychological tests used on humans to explore whether powerful AI chatbots, such as GPT-3.5, have developed the ability to infer the thoughts of other humans, a capacity known as the theory of mind (ToM). The tests consisted of giving the AI systems a scenario and asking questions to test their understanding of what was happening, including testing their comprehension of characters’ expectations, beliefs, etc. Kosinski’s results revealed that the most recent AI models performed very well on such tasks, comparable to the level of seven and nine-year-old children. When the AI becomes equipped with the ToM, it will likely master or at least comprehend social interactions, communication, empathy, self-consciousness, and morality, all of which are critical in human negotiation. [References: 1, 2]
So, what do all of these developments mean for the future of our society? AI can potentially transform the world in significant ways, but it’s essential to consider the ethical implications and develop clear guidelines and standards for its use. That’s where the Blueprint for an AI Bill of Rights (Blueprint) comes in. The Blueprint is a framework released by the White House Office of Science and Technology Policy in October 2022. Here are the five principles quoted from the Blueprint:
Safe and Effective Systems: You should be protected from unsafe or ineffective systems.
Algorithmic Discrimination Protections: You should not face discrimination by algorithms and systems should be used and designed in an equitable way.
Data Privacy: You should be protected from abusive data practices via built-in protections and you should have agency over how data about you is used.
Notice and Explanation: You should know that an automated system is being used and understand how and why it contributes to outcomes that impact you.
Human Alternatives, Consideration, and Fallback: You should be able to opt out, where appropriate, and have access to a person who can quickly consider and remedy problems you encounter.
The Blueprint provides a framework for AI guidelines when a federal agency lacks specific guidelines for the industry it supervises or its affairs. While it’s currently a non-binding framework, it offers a set of principles for developing AI guidelines and standards in the future. [Reference: 1; Supplement: Opportunities and blind spots in the White House’s blueprint for an AI Bill of Rights]
That’s all for this week. I will see you next Thursday!
One Question This Week:
What excites or scares me the most about the future of artificial intelligence?
Stay safe and sharp,
Jason Lai
Disclaimer: Not legal advice. You can see the full disclaimer & disclosure here.
Credit: Originating from my prompts, research, and reflection on AI, this Brief was reviewed and edited by ChatGPT.