One of the human’s challenge is negotiation. We know what we want, what is best for us, better yet, we know “which is the fairest option”. We expect the other side to know the same thing but if we were to lay the cards on the table we fear that they may benefit from the information. Therefore, “hiding some information” and “exposing only what we want to show” while trying to move towards mutual benefit (or “acceptable” results) are the challenges in negotiation. Some go so far to say that negotiation has no final answer, hence, the phrase the “art of negotiation”.
Not everyone is born with the ability to negotiate. Even with the smartest individuals, once facing another person, their negotiation plan can fail. This may be why we need lawyers or representatives to help us negotiate. This is because once negotiation is done via a “representative” (who seem to be third-party), factors such as “manners” or “fear” seem controllable.
Computers are gradually becoming negotiation representatives. Looking beyond “large” negotiations such as international negotiations on carbon, company trade deals, or other negotiations for tangible benefits, we see that “small” negotiations such as communication for traffic lights “queues” or for solar cell panels and electricity (price negotiations) can be enhanced for higher efficiency.
At the International Joint Conference on Artificial Intelligence or IJCAI in Melbourne, Australia last month, AI researchers from the Netherlands, UK, and the USA jointly presented a research named When Will Negotiation Agents Be Able to Represent Us? The Challenges and Opportunities for Autonomous Negotiators. This research explored the challenges, development opportunities, and future social benefits of autonomous negotiators.
The team believes that human and some organization avoid negotiation out of fear or lack of skill. The weakness contributes to social inequality, political gridlock, and social injustice. They mention that automated negotiation can offer better win-win deals, reduce time, costs, stress and cognitive effort on the part of the negotiator. Moreover, they say that this is not a luxurious option but will be a requirement in instances where a human decision is too slow and costly. For instance, in the future when the smart electrical grid is deployed worldwide, household devices shall be able to automatically negotiate and renegotiate complex energy contracts.
The team suggested 3 challenges for automated negotiations that are awaiting development.
- Domain knowledge and preference elicitation – The automated negotiators need to learn adequate information from the user to construct the accurate preference model. Before the negotiation, the automated negotiator must know “what we prefer” and although they may not know exactly what we prefer, they must know what we value.
Tim Baarslag, head researcher, explained to Science magazine about a job negotiation that the system might suggest that “From the proposals received so far, it can be deduced that the salary level is very important for your boss, while you have previously indicated you value family life most of all. Perhaps asking instead for a work day at home could be in your mutual interest.”
- Long-term perspective – Automated negotiators need to learn that negotiations are not one-time and will have repeated encounters. For example, with a smart home where different occupants will have different needs and preferences and have to reach mutual agreements for instances such as the temperature of the house and the use of devices.
For this matter, the automated negotiator needs to incorporate history and future as a factor (the researcher admits that it is uncertain and is difficult to seek a common stance). The researchers suggest an agreement such as “for free” or “agree for a disadvantage” if the system has a “reputation system” or “credibility score” (which is similar to the human decision-making process).
- User trust and adoption – The researchers suggest that to gain trust, automated negotiators must incorporate 3 factors, “user participation”, “transparency”, and an “appropriate communication system”.
For “user participation”, they suggest that the system should “do the brunt of the negotiation work to find possible agreements with the other negotiation parties and which can present to their human partners for feedback and new input”. For “transparency and communication”, the automated negotiator should provide the user with the factors behind each decision.
Aside from these challenges, the conclusion raises concerns that, normally, in negotiations via third (human) parties, people may show less regard for fairness and ethical behavior. Therefore, with automated negotiators, such concerns may also arise. Researchers suggest solutions such as having agreements beforehand to state that users are responsible for the consequences based on the settings done by the user.
Despite the challenges and problems, Tim Baarslag believes that automated negotiators will prove beneficial. He mentions that we may be using these negotiators from small-scale matters such as online trade, property trade, setting a meeting, or unlocking political gridlock when parties need to negotiate deals such as the Paris agreement.
Digital Ventures x Champ Teepagorn
Champ Teepagorn from ‘Tempura Culture’, Thai PBS.
Author of World While Web in A Day Magazine and Head in the clouds in GM Magazine.
Good at writing, like drawing comics and illustrations.