| Project/Area Number |
18H03300
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 61030:Intelligent informatics-related
|
|---|
| Research Institution | National Institute of Advanced Industrial Science and Technology |
|---|
Principal Investigator |
Kamishima Toshihiro 国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 主任研究員 (50356820)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
馬場 雪乃 筑波大学, システム情報系, 准教授 (40711453)
鹿島 久嗣 京都大学, 情報学研究科, 教授 (80545583)
|
|---|
| Project Period (FY) |
2018-04-01 – 2021-03-31
|
| Project Status |
Completed (Fiscal Year 2021)
|
| Budget Amount *help |
¥10,660,000 (Direct Cost: ¥8,200,000、Indirect Cost: ¥2,460,000)
Fiscal Year 2020: ¥3,380,000 (Direct Cost: ¥2,600,000、Indirect Cost: ¥780,000)
Fiscal Year 2019: ¥4,030,000 (Direct Cost: ¥3,100,000、Indirect Cost: ¥930,000)
Fiscal Year 2018: ¥3,250,000 (Direct Cost: ¥2,500,000、Indirect Cost: ¥750,000)
|
|---|
| Keywords | 公平性 / クラウドソーシング / 機械学習 / データマイニング |
|---|
| Outline of Final Research Achievements |
We tackle with one task related to fairness-aware machine learning, in which the predictor tries to satisfy some fairness constraint, such as statistical stability. In the previous literature of fairness-aware machine learning, the prediction accuracy is evaluated on the observed dataset whose decision labels are supposed to be unfair. This is due to the restriction that truly fair labels cannot be observed. We tried to evaluate the precision on the fair labels as proxy by using preference data influenced by cognitive biases. We collected such data through a crowdsourcing service. Then, to evaluate how much information of fair decisions are extracted from these observations, we developed the notion of stability and a method to quantify the stability.
|
| Academic Significance and Societal Importance of the Research Achievements |
機械学習の公平性については2011年から取り組んでいるが，2016年の米大統領選や，欧州のGDPR試行に伴い注目され，世界的に研究が拡大している研究分野である．しかしながら，本当はあるべき公平な決定というものが観測できない根本的な制限がある．この制限に対して，センシティブ情報の代用として認知バイアスを利用して，人工的にデータと収集するという手段で挑んだのが本研究である．
|
