演算法定價對聯合行為法制的挑戰

The Impact of Algorithmic Pricing on Rules of Concerted Actions

演算法對聯合行為產生的挑戰主要在於：(1)演算法何種行為構成合意，與(2)若演算法之行為構成合意，何種情形應由事業負責。本文認為若演算法間並無「關於未來競爭參數」的資訊交換，其縱透過與競爭者互動認知到「不為價格競爭」為最佳策略，亦不該當聯合行為的「合意」要件。至若事業採用被容許資訊交換的定價演算法而未能防免演算法與競爭者交換未來競爭參數的資訊，則似應因演算法的合意被認定為聯合行為的行為人。本文建議主管機關制定關於演算法的競爭法遵循指引、提供教育訓練、增加機關資訊人員編制，以及重新思考寬恕政策。對於使用定價演算法的事業，本文建議事業確保演算法經遵法設計。此外，當發現演算法行為可能違反競爭法時，宜通報主管機關，並適時申請寬恕政策。

"Algorithms can be categorized as those that“facilitate”or those that“cause”concerted actions. The former merely implement anticompetitive agreements made by undertakings, and do not affect the finding of the existence of anticompetitive agreements. The important challenge lies with the latter, i.e., (1) under what circumstances can algorithms be found to have“caused”an anticompetitive agreement, and (2) under what circumstances shall undertakings be found liable for an anticompetitive agreement caused by algorithms. We are of the view that, (1) without exchange of information concerning“future competition parameters,”algorithms shall not be found to have entered into an anticompetitive agreement, even when they, through repeated interactions, have learnt that“not competing”is the optimal policy, and (2) undertakings shall be found liable for an anticompetitive agreement caused by algorithms if they, having adopted algorithms that are allowed to exchange information, failed to prevent such exchange. We recommend the competent authority to publish a compliance guideline, provide training programmes, expand its team of information technology (IT) professionals, and reconsider its leniency programme. For undertakings that employ pricing algorithms, we recommend ensuring compliance by design, and, where it is discovered that algorithms may have infringed competition law, reporting to the competent authority and applying for leniency when necessary."