http://dx.doi.org/10.6035/recerca.5516 Polisの論文
Deliberative and participatory approaches to democracy seek to directly include citizens in decision-making and agenda-setting processes. These methods date back to the very foun-dations of democracy in Athens, where regular citizens shared the burden of governance and deliberated every major issue. However, thinkers at the time rightly believed that these meth-ods could not function beyond the scale of the city-state, or polis. Representative democracy as an innovation improved on the scalability of collective decision making, but in doing so, sacrificed the extent to which regular citizens could participate in deliberation. Modern tech-nology, including advances in computational power, machine learning algorithms, and data visualization techniques, presents aunique opportunity to scale out deliberative processes. Here we describe Polis, an open source web application capable of collecting and synthesizing feedback from people in a scalable and distributed fashion. Polis has shown itself capable of building shared understanding, disincentivizing counterproductive behavior (trolling), and cultivating points of consensus. It has done this in the context of journalistic and academic research, and directly as part of decision-making bodies at local and national levels, directly affecting legislation. These results demonstrate that deliberative processes can be scaled up beyond the constraints of in-person gatherings and small groups.Key Words: deliberation, collective intelligence, unsupervised learning, active learning.
欠損値はコメントごとの平均で埋められる 投票の少ないユーザ(<7)は取り除かれる PCAはパワーイテレーション法で計算する、前回の計算結果を再利用できるので効率的
欠損値を平均で埋めると、投票数の少ない人が中央付近にプロットされることになる
意見グループ
各コメントがグループをどの程度代表してるか
p.9の上半分はスキップ
ユーザの時間を効率よく使うためにコメントは重み付きランダムで選ばれる 優先度の式は ・賛成が多いと増える項 ・パスが多いと減る項 ・PCA負荷の高いコメントで増える項 ・投票が多くなるほど減る項 を掛けたもの