Table 2 Hypotheses and findings overview... | Emerald Publishing

Table 2

Hypotheses and findings overview

Studies	Hypotheses	Findings	Attribution patterns for success and failure
Study 1a	H1a: Managers attribute successful AI (vs. human) performance to external (internal) factors H1b: Managers attribute unsuccessful AI (vs. human) performance to internal (external) factors	Managers attribute the success of AI agents to external factors unrelated to the AI’s characteristics. On the contrary, the success of a human Agent is attributed internally, i.e. to the Agent’s characteristics. Thus, H1a is supported We did not observe a statistically significant difference between humans and AI in a failure condition. Thus, we reject H1b	If a human manager succeeds with an AI agent, the credit goes to the human manager; if the manager succeeds with a human agent, the credit goes to the human agent, not the manager. However, in the case of failure, the attribution of responsibility does not differ based on the type of agent
Study 1b		Considering the difference in causal attributions of successful vs. unsuccessful AI during a real interaction, we observed that the success of AI is attributed externally to itself (managers take credit for the success). In contrast, the failure of AI is attributed internally (managers blame the algorithm). Thus, we partially support H1a and H1b as Study 1b did not consider conditions involving human agents	If it’s successful interaction, the credit is mine; but if it’s unsuccessful, the fault lies with the AI, not me
Study 2	H2: Perceived locus of causality mediates anticipated satisfaction with the service agent so that external attribution of successful AI agent is associated with lesser anticipated satisfaction	Managers expect less satisfaction from successful AI vs. humans as they perceive the success of AI externally as not related to its characteristics. By that, we support the mediative role of locus of causality in anticipated satisfaction from the Agent	In a successful AI Agent-manager interaction, the human manager takes credit, not the Agent; but in human agent-manager interaction, success is attributed to human agent, not the manager

Studies	Hypotheses	Findings	Attribution patterns for success and failure
Study 1a	H1a: Managers attribute successful AI (vs. human) performance to external (internal) factorsH1b: Managers attribute unsuccessful AI (vs. human) performance to internal (external) factors	Managers attribute the success of AI agents to external factors unrelated to the AI’s characteristics. On the contrary, the success of a human Agent is attributed internally, i.e. to the Agent’s characteristics. Thus, H1a is supportedWe did not observe a statistically significant difference between humans and AI in a failure condition. Thus, we reject H1b	If a human manager succeeds with an AI agent, the credit goes to the human manager; if the manager succeeds with a human agent, the credit goes to the human agent, not the manager. However, in the case of failure, the attribution of responsibility does not differ based on the type of agent
Study 1b		Considering the difference in causal attributions of successful vs. unsuccessful AI during a real interaction, we observed that the success of AI is attributed externally to itself (managers take credit for the success). In contrast, the failure of AI is attributed internally (managers blame the algorithm). Thus, we partially support H1a and H1b as Study 1b did not consider conditions involving human agents	If it’s successful interaction, the credit is mine; but if it’s unsuccessful, the fault lies with the AI, not me
Study 2	H2: Perceived locus of causality mediates anticipated satisfaction with the service agent so that external attribution of successful AI agent is associated with lesser anticipated satisfaction	Managers expect less satisfaction from successful AI vs. humans as they perceive the success of AI externally as not related to its characteristics. By that, we support the mediative role of locus of causality in anticipated satisfaction from the Agent	In a successful AI Agent-manager interaction, the human manager takes credit, not the Agent; but in human agent-manager interaction, success is attributed to human agent, not the manager

Source(s): Table by authors