Cooperative AI: machines must learn to find common ground

“…As found in 52 in a saturated network case a distributed agent scheme can beat the traditional central allocation. As one step further, cooperative AI‐agents can surpass the performance of non‐cooperative agents, 53 which can be proved in this use case well.…”

“…These positive results are in line with research by Gogola, Dębski, Goryczka, Gorczyca, & Piegza, (2021) which states that suitability or compatibility is an important aspect of fit (ex ante) and will influence the extent to which partners are able to realize the synergistic potential of collaboration. Furthermore, Dafoe et al, (2021) found that compatibility can encourage commitment and communication processes through shared goals -which in this research is defined as strategic alignment where compatibility in this research is viewed from the perspective of corporate culture and appropriate operations strategy (Barney et al, 2021). However, it is further emphasized that compatibility does not mean that it is generic or applied equally but rather refers to mutually agreed interests.…”

“…Such an interface would allow designers to express their intentions in natural language and understand the system's reasoning behind its suggestions, thereby guiding the optimization process more effectively-capabilities that are often limited in standard BO, where the system leads the process with minimal room for intervention. This design aligns with the principles of cooperative AI [6], which emphasize shared understanding, communication, and commitment in human-AI collaboration.…”