CYIL vol. 11 (2020)

CYIL 11 (2020) ARTIFICIAL INTELLIGENCE AND COMPETITION LAW… In perfectly competitive markets, such uncoordinated conduct may indeed have no anticompetitive effects. The situation may however be different on oligopolistic markets, where the market power of individual undertakings guarantees that – unlike under perfect competition – their individual conduct has a perceptible influence on market output. 61 In order to behave rationally, one oligopoly therefore needs to react to the conduct of the others. This does not however mean that they need to “follow” each other. If one of the undertakings increases its prices, it may indeed by a rational reaction for the other one to decrease its prices, and thus acquire more customers. The game theory, which is often used to understand the functioning of oligopolistic markets, calls such situations “a game” and categorizes such reactions as “cooperate” and “defect”; to put it simply, if one oligopoly increases its prices, the other may either “cooperate”, e.g. increase its prices as well, or “defect”, i.e. keep its prices intact or even lower them. Imagine a market with two oligopolistic “players”, who may either increase their price (high price) or decrease it (low price). If both of them opt for low price, their profits will be relatively low, close to perfectly competitive markets. If, on the other hand, they choose the high price, their profits will increase to levels close to monopolistic markets. If, however, one “player” increases the price and the other reacts by decreasing it, the customers would follow the low price, securing significantly higher profits for the latter and very low ones for the former, as is described in the table below: 62 Player 1 4, 4 In a single game, it is therefore rational for the second player to defect, i.e. to opt for a low price and thus maximise its profits; if, however, the number of games increases, the reaction of the first player would predictably be to lower its prices as well, which will leave them both with low profits. On the other hand, if the second player choses to cooperate, i.e. to increase its prices as well, it will end up with a smaller profit in the first game, but in a longer term, both of them would keep higher prices and thus secure higher profits. Even though the players cannot communicate directly, they can “learn” frompast experience, which becomes more and more important as the number of games grows. This is explained by the so called reward-punishment scheme: if one player “cooperates” and increases its prices, it will be “rewarded” by the other, who will increase its prices as well; conversely, if a player “defects” and lowers its price, it will be “punished” by a price war. 63 Thus, the intention to maximise short-term profits by “defection” leads to “price wars” and lower long-term profits, whereas “cooperation” leads to higher profits in the future. The market price may therefore be increased in this way by individual, non-coordinated, but mutually reactive conduct of the oligopolies; the economic theory calls this phenomenon a tacit collusion: 61 FAULL, NIKPAY ( op. cit. sub 55), p. 25. 62 HARRINGTON, J. E. Developing Competition Law for Collusion by Autonomous Artificial Agents. Journal of Competition Law & Economics , 2019 (3), p. 343. 63 Ibid , p. 355. Player 2 Low price High price Low price High price 2, 2 1, 6 6, 1

249