Computers can play chess but can they negotiate business deals?

 

The game of chess and the process of business negotiation share many similarities. Two sides engage in a strategic dance toward their objectives. Back and forth, each player's move affects the other's next move in an exciting tangle of calculation and strategy.

As technology has advanced, computers have been learning how to play human games. IBM's Deep Blue beat world chess champion Garry Kasparov in a six-game match in 1997. Unlike chess, however, business negotiations involve the deeply human elements of collaboration, emotion, language, subjectivity and trust, all of which have to be taken into account. That's why few would have anticipated that, only 15 years after Deep Blue's victory, computers would be playing a bigger role in the much more complex game of negotiation.

Nowadays computers can promote win-win strategies and even trust in online sales negotiations. For a recent paper, which I wrote in collaboration with Yinping Yang of A-Star, Nuno Delicado of Pluris and Andrew Ortony of Northwestern University in Evanston, Ill., we found that trust can be built between humans and computers by adding a simple dynamic into the mix: taking the initiative of putting a single priority on the table, explaining the motivation to do so and inviting one's counterpart to do the same.

While face-to-face negotiations can be of benefit if the individuals involved trust each other, trust also is important in online interactions. Our experiments suggested that, by volunteering information that it need not disclose, a computer agent can alleviate mistrust in humans engaging with it.

We know that, in human-to-human negotiations, if a win-win negotiation move is adopted, such as proactively sharing interests, this can yield more value. There are many advantages to win-win strategies: long-term business relationships, efficient processes and more value in the outcomes for both sides. What is fascinating in our findings is the discovery that what works in human-to-human negotiations also seems to work in computer-to-human negotiations. These findings have practical implications for companies using software in negotiations.

We conducted a multi-issue negotiation in which a computer agent was the seller and humans the buyer of laptop computers. The machine had four issues in its negotiation arsenal: price, quantity, service level and delivery terms. In one condition the computer honestly revealed its No. 1 priority, price. In this condition, however, even if the human counterparts revealed their preference back, the computer did nothing to maximize the preferences of the human counterpart. Interestingly, the perception among the human participants was that it did.

There was a marked difference in the number of agreements when the computer was proactive in sharing its priority, with 22 out of 27 possible agreements, compared to 14 out of 27 when it was not. Similar results were reflected in the satisfaction of the "buyer." The majority of participants also responded to the computer's invitation to share their priorities to align with its four issues.

Even more interesting was the discovery that distrusting humans came on board with the machine once it put one of its cards on the table, shared its intention to collaborate and invited the other party to reciprocate. In this case "Machiavellian" personality types, who are less trusting, reacted similarly to those with more trusting personalities during the negotiation. This suggests that, if you make the right moves and share information that can help both parties become better off throughout the negotiation, you can normalize even distrustful counterparts. This cuts the need to try to profile your opponent before a negotiation.