Human-Agent Negotiation League (HAN)
Part of the ANAC Competition

Details of HAN 2025 Pilot can be found here. See you in HAN 2026

The Human-Agent Negotiation League (HAN) is concerned with advancing the state of the art in human-agent negotiation technology. In 2025, a pilot competition was conducted as part of ANAC 2025 where participants of IJCAI 2025 were invited to compete as human negotiators against a simple heuristic strategy. We plan to release anonymized logs of these negotiations as a public dataset to be used for informing the design of new automated agents capable of effective negotiation with human partners.

Future versions of HAN will focus on competitions between agent designers as well as negotiation support tool designers. Stay tuned.

  • August 22nd, 2025 21:14 EDT The winners of HAN 2025 Pilot were announced at the closing session: Hossein Savari, University of Tehran and Shota Takayama, Tokyo University for Science and Technology. They will split the 1000USD prize equally.
  • August 22nd, 2025 14:46 EDT If you are a participant of HAN 2025, please fill the Post-Competition-Questionnaire.
  • August 22nd, 2025 14:33 EDT HAN 2025 Pilot is concluded. We will announce the winner(s) in the closing session of IJCAI 5pm today. We thank all participants for their great efforts. The leaderboard will be shown again after the conclusion of IJCAI 2025. Moreover, we plan to re-open the game for everyone in the next few days (but for no prizes this time).
  • August 22nd, 2025 09:30 EDT Due to high traffic at some time this morning, some participants may have experienced sudden failure of negotiations. This is dealt with now.
  • August 21st, 2025 17:34 EDT We found an issue for people conducting more than 100 negotiations, they would see repetitions of the same scenarios in some cases. This was fixed now.
  • August 21st, 2025 6:28 EDT The UI just restated. If you were engaged in a negotiation exactly at this time, you may have experienced a sudden drop of the negotiation. Rest assured that this did not affect your score in any way.
  • August 21st, 2025 6:11 EDT The reward announcement will be on Friday August 22nd after 3pm. You do not need to attend this ceremony in person in order to win a prize. We will contact the winners by email after the conclusion of IJCAI to arrange the transfer of their prices.
  • August 20th, 2025 18:00 EDT The agent utility display bug is resolved
  • August 20th, 2025 13:17 EDT We discovered a display bug in the "User Results" tab, The agent utility is reported as the human utility. This does not affect scores in any way and it is only on the display. The data is correct. You can instead see the welfare field which is the sum of agent and human utility. To avoid disrupting running negotiations, we decided not to apply the fix immediately. The fix will be applied today when we notice no running negotiations.
  • August 20th, 2025 09:00 EDT The competition started!! The official competition ends Friday, August 22nd 12pm EDT (last day of IJCAI).

In this pilot for the Human-Agent Negotiation league, participants of IJCAI 2025 will compete against automated negotiation agents in various negotiation scenarios. Winners will be announced on the last day of IJCAI 2025 (Friday: August 22nd, 2025). The winners will split a total of up to 1000USD prize pool.


To help us analyze the results of this competition, we ask all participants to fill the following questionnaires (each requiring only a couple of minutes):

To participate, you need to follow the following steps:

  1. Drop by the ANAC booth anytime between 9am and 12am or 2pm and 5pm to register an account (best during coffee breaks). You can also register an account online here
  2. After registration, you will be able to log into the system and start negotiating using the Human Agent Negotiation Interface.
This year's competition will use a simple bargaining (Alternating Offers) protocol. When you start a negotiation, the AI agent will present you with their offer. You can either (1) accept it, ending the negotiation with success receiving your utility for the offer which is now the agreement. (2) end the negotiation in failure receiving the reserved value. (3) reject the offer and send a counteroffer, continuing the negotiation. The AI agent will then have the same 3 options. This process continues until either (1) you or the AI agent accept an offer, or (2) you or the AI agent end the negotiation, or (3) 5 minutes have passed without agreement, (4) more than 100 offers have been exchanged without agreement. In the last two cases, the negotiation ends with failure and you received your reserved value.

We provide some resources to help you understand the negotiation interface and be ready to participate in the competition:

  1. This video describes how to use the interface.
  2. If you prefer text, a brief explanation of the negotiation system can be found here.

You must be 18 years or older to participate in this condition, Furthermore, you must satisfy the following conditions to be considered for the official competition and the monetary prize

  1. You must agree to the consent form during registration (here).
  2. You must fill the Pre-Competition-Questionnaire before conducting any negotiations, and the Post-Competition-Questionnaire after conducting all negotiations.
  3. There are three different negotiation scenario types in HAN 2025 (Grocery, Island and Simple Trade). You need to log at least two negotiations on each scenario.

When calculating the score of each participant we will follow the following procedure:

  1. The lowest score on each scenario type will always be discarded.
  2. If you record more than 4 negotiations on any scenario type, all scores other than your top 3 will be discarded.
  3. For each scenario type, the average score on negotiations that were not discarded will be calculated as the scenario type score.
  4. Your final score is the average of scenario type scores. This final score is guaranteed to be a number between zero and one.
  5. Discarded scores (except the lowest per participant) may be used for tie-breaking if two participants have almost the same final score (i.e. up to two significant digits).

The winners are the participants with maximum final scores. The organizers will decide the number of winners based on the number of participants and score distribution but it will not exceed three winners. A prize pool of up to 1000USD will be divided between the winners.

The winners will be contacted after the conference to arrange the transfer of their prizes. This process may require some time to finalize.

ANAC runs every year with different challenges. Keep an eye on this website for future competitions. If you are interested in developing automated negotiating agents to negotiate with people or other agents using structured protocols or natural language, you are in the right place.

To receive updates and to prepare for participating in future ANAC competitions, please register.

You can find more about automated negotiation and the NegMAS platform underlying the competition here (This is strictly optinoal. You do not need any of this for the HAN competition.):

  1. For a simple introduction to the automated negotiation field check this tutorial
  2. For more information about developing automated negotiating agents, you can check NegMAS Documentation
  3. If you prefer a book check this recently published open-source book.

Links

Important Dates (EDT)

First DayAugust 20
Last DayAugust 22
Competition ClosesAugust 22, 2pm EDT
LeaderBoard Update StopsAugust 21, 7pm EDT
Winner AnnouncementAugust 22 after 3pm

Previous Competitions

Human Leader Board

#UserScore N
1 Hossein 95.6383 177
2 shota 95.3737 203
3 David 86.3452 154
4 atsunaga 84.5014 171
5 Tianzi Ma 82.6123 119
6 mh 80.1222 62
7 Tarek 79.31 74
8 witsyke 78.2269 72
9 furkancanturk 76.8016 76
10 ATrue 74.4271 22
11 HenryF 72.3138 54
12 Amina Anna 72.0549 73
13 fensan 70.5042 49
14 ekiral 67.5332 26
15 Klone 61.5628 8
16 Ebrahim Bagheri 60.5076 19
17 berkecankocyigit 43.7556 20
18 mpsingh 39.7635 7
19 royyang 35.88 9
20 Pascaliversen 33.794 15
21 Yadjm084 29.6162 10
22 WasayS 29.3289 6
23 monurkeskin 27.4672 13
24 Faezeh Ensan 1.5099 2
25 aarikan 0 2
26 emre.kuru@ozu.edu.tr 0 1
27 SteinStein 0 1

Statistics

Number of negotiations: 1452

Number of participants: 27

Scenario N Mean Top Agreement
Trade 446 36.45 100 49%
Island 477 30.93 93.9394 47%
Grocery 529 29.75 97.8723 40%