CSCE 496/896

Handout 9: Negotiation Day Analysis

November 26 2002

Round 1: Discussion of Results

Phase I:

Groups	# Unique Products	Money	Bingo?	Score
The Duke	6	$450	5	0.65
Tod & Copper	6	$500	2	0.83
Koala	6	$450	6	0.60
Runner	6	$500	4	0.73
Trackers	6	$600	3	0.85
Super Hunter	6	$500	1	0.88

Table 1 Results of Round 1 Phase I.

The score is 0.6*(#unique products + (6-Bingo))/12 + 0.4*(Money/maximum). The maximum value is $600. The winner for this round and phase is Super Hunter, followed by Trackers and Tod & Copper. Note that Trackers is also the only team that ended up with more money than the original amount. That was good work.

Phase II:

Groups	# Unique Products	Money	Bingo?	Score
The Duke	6	$0	4	0.40
Tod & Copper	1	$2500	NA	0.40
Koala	6	$0	2	0.50
Runner	6	$0	3	0.45
Trackers	5	$500	NA	0.08
Super Hunter	6	$0	1	0.55

Table 2 Results of Round 1 Phase II.

The score is 0.6*(#unique products + (6-Bingo))/12 + 0.4*(Money/maximum). The maximum value is $2500. For Tod & Copper, Since their Bingo is infinite, they receive 0 points for the 0.6 portion of the score. The winner for this round and phase is Super Hunter, followed by Koala and Runner. Note that Tod & Copper executed a monopoly in this phase and they ended up with $2500 and broke the banks of four other teams. Note also that Trackers tried to form an alliance against the monopoly of Tod & Copper but failed; and they failed to obtain 6 unique products. As a result, they suffered the most. But that was a valiant attempt. Unfortunately, it did not work.

Groups	Phase I	Phase II	Average
The Duke	0.65	0.40	0.525
Tod & Copper	0.83	0.40	0.615
Koala	0.60	0.50	0.55
Runner	0.73	0.45	0.59
Trackers	0.85	0.08	0.465
Super Hunter	0.88	0.55	0.715

Table 3 Results of Round 1, average.

The Winner of Round 1 is Super Hunter, with an average score of 0.715, followed by Tod & Copper, Runner, Koala, The Duke, and Trackers. But the effort by Trackers is commendable.

Round 2: Hostage Rescue

Situation 1: School Children, Koala the police, Tod & Copper the Kidnappers.

Situation 2: Forest Fire, Runner the police, The Duke the anti-environmentalists.

Situation 3: Oil Tanker, Trackers the police, Super Hunter the Earth protectionists.

Groups	Appeal Prevailing Practice	Counter-Example	Appeal Past Promise	Appeal Self Interest	Promise Future Reward	Threat	Score (#types/10)
The Duke	Yes	Yes	Yes	Yes		Yes	0.5
Tod & Copper	Yes	Yes			Yes	Yes	0.4
Koala			Yes	Yes	Yes	Yes	0.4
Runner	Yes	Yes	Yes	Yes	Yes	Yes	1.0
Trackers		Yes	Yes	Yes	Yes	Yes	0.5
Super Hunter		Yes	Yes	Yes			0.3

Table 4 Results of Round 2 Argumentative Negotiation in Hostage Rescue Situations.

The winner of this round is Runner, followed by Trackers and The Duke. The effort by Runner is very commendable. Runner’s arguments were methodical, logical, and very in context. That was good effort. Also, both Trackers and Super Hunter did very well: their arguments were very much in context, logical, and very sensible. It was unfortunate that they came to a deal and could not go through all argument types.

Groups	Round 1	Round 2	Total
The Duke	0.525	0.5	1.025
Tod & Copper	0.615	0.4	1.015
Koala	0.55	0.4	0.95
Runner	0.59	1.0	1.59
Trackers	0.465	0.5	0.965
Super Hunter	0.715	0.3	1.015

Table 4 Results of game day, total.

The winner of the Game Day is Runner, followed by The Duke, and then by Super Hunter and Tod & Copper (tied for the third place), and then by Trackers, and finally by Koala.

Team Observations

1. The Duke: This team came in with a pre-game strategy. They made very detailed and good in-game observations.

2. Tod & Copper: This team came in with a set of pre-game strategies. This team also made some good observations. They were also the team that exercised complete monopoly in Round 2 Phase II.

3. Koala: This team came in with a set of pre-game strategies. They also made good in-game observations. In addition, their pre-game negotiation strategies for Round 2 were very detailed.

4. Runner: This team came in with a set of pre-game strategies. Amazingly, with only one member present, this team was able to BINGO (accomplish its task) first in Round 1 Phase I. This team had a very good Round 2. This team did not make enough in-game observations.

5. Trackers: This team came in with a set of pre-game strategies. Their strategies were well-thought out and they had good questions about the setup of the game. The also made good in-game observations about other teams. They were very good at executing their monopoly power (partially) in Round 1 Phase I. However, they failed to form an alliance against Tod & Copper (the monopoly) in Round 1 Phase II. It was a valiant effort.

6. Super Hunter: This team came in with a set of pre-game strategies but did not make enough in-game strategies.

Conclusions

1. For Round 1, there was a significant difference in the teams’ behavior in Phase I and Phase II. Because of a middle person in Phase I, the teams were conservative. They opted to swap goods to begin with. Why? It was inconvenient for a team to execute monopoly power due to the communication delay. It was difficult for a team to check out the supply and demand in the market and thus difficult for a team to raise the price of its products. In Phase II, each team was able to gather more information more quickly and had more control over the negotiations. Thus, this phase is more convenient for a team to execute its monopoly power. So, even though in both phases, each team had the monopoly power, only one team in Phase II exercised it fully (Tod & Copper) and another team in Phase I exercised it partially (Trackers).

Lesson Learned: Even in a setup like the above, agents can still cooperate fairly because of the reliance on each other for products/services to accomplish own tasks. This gives us designers great flexibility in designing our agents. We can give our agents unique characteristics and be competitive and self-interested, and we can still guarantee to some extent that the agents will cooperate.

2. In Round 1 Phase I, there were teams that worked faster than the other teams. And teams that worked fast (processing offers, processing requests, making offers, etc.) were able to accomplish their tasks sooner. In addition, they were able to raise their products’ prices after they had accomplished their tasks.

Lesson Learned: So this tells us that in a dynamic, real-time environment, agents that work fast are more advantageous.

3. In Round 2, there were teams who planned well and were methodical, logical, and in context. This was very good to see. The idea was to how to apply those six argument types into the situation well. There were teams who were creative and imaginative. Very good.

4. When we design agents and use negotiations as the primary interaction protocol, we are motivated by mainly two things: (a) the need for negotiations, and (b) the benefits of using negotiations. In our games, the need for negotiations was evident. To accomplish tasks, each agent needs to cooperate with other agents. And since each agent also wants to maximize its own benefits, it wants to make sure that every time it agrees to help another agent, it gets something back in return. So, very easily, we can motivate the agents to negotiate without imposing any more global instructions! Secondly, what is the benefit of using negotiations? Agents have different goals and not all agents are readily willing to help. So, negotiations bring agents to a deal, to a joint goal or intention. Without negotiations, agents cannot arrive at a deal. Without negotiations, an agent cannot try to maximize its benefits. Without negotiations, agents cannot recover from bad deals. So, there is a need for negotiations, and on top of that, agents can improve their performance through negotiations. In other words, negotiations allow agents to solve a problem, and also to solve a problem better.