Algorithmic and Human Collusion

Abstract

The working papers published in the series constitute work in progress circulated to stimulate discussion and critical comments. Views expressed represent exclusively the authors' own opinions and do not necessarily reflect those of the editor.

Algorithmic Price Recommendations and Collusion: Experimental Evidence

“…Johnson et al (2020) focus on tacit collusion among self-learning algorithms on sales platforms and discuss how the platform's design choices influence it. and Werner (2022) show experimentally that algorithms may raise market prices even above the price level usually observed in human markets. We differ from this approach as we consider algorithms that only give recommendations but do not compete with the other firms in the market.…”

“…Wieting and Sapi (2021) and Musolff (2022) show that real-world pricing algorithms are often rule-based and follow straightforward conditional processes. Moreover, although alternative methods like reinforcement learning algorithms have more complex routines to learn a pricing strategy, they eventually often converge to strategies that simple rules can describe (Werner, 2022;Klein, 2021;Kasberger et al, 2023). 11 Hence, our focus on those algorithms is attractive from a methodological perspective and realistic regarding the tools used in actual markets.…”

“…Recent studies byMusolff (2022) andWieting and Sapi (2021) highlight that price algorithms on sales platforms often follow simple rule-based logic. Similarly, even complex reinforcement learning algorithms often converge to strategies that simple rules can describe (see, for instance,Werner, 2022).…”

Will the use of machines to identify negative activist investments backfire? Experimental evidence

“…Here, we investigate what happens when machines are on the well-intentioned "prosecution team", that turns out to make a harmful false-positive error (Rebitschek et al, 2021). Thus, in contrast to a considerable line of research focusing on how malevolent actors use algorithms to their benefit (Acemoglu, 2021;Köbis et al, 2021;Werner, 2021), we focus on what happens when benevolent actors use them. By doing so, we intend to uncover the envelope of acceptable human-machine collaboration (Haesevoets et al, 2021;Rahwan, 2018).…”

Algorithmische Preissetzung und tacit collusion

“…716 M.Nowak/Sigmund, Nature 364 (6432) (1993), 56. 717Werner (2021), S. 28. 718Siehe Calvano et al (2020), AER 110 (10), 3267 (3288).719 Calvano et al (2021), IJIO 79, 102712. den Algorithmen zur Verfügung stehen.720 Während das Verhalten der Algo rithmen bei perfekter Überwachung dem festgestellten Verhalten aus den Bertrand-Simulationen 721 entspricht, nimmt die Kollusion bei Wegfall der Informationen ab, wenngleich sie mit 75% des perfekten Kollusionsgewinns noch immer deutlich über den nicht-kooperativen Gleichgewicht liegt.722 Algorithmische Kollusion mittel deep learning Algorithmen Vorteile gegenüber Q-Learning Q-learning Algorithmen sind relativ unkompliziert und können "mit wenigen Parametern, deren ökonomische Bedeutung klar ist, vollständig charakteri siert werden."…”