In recent years, the realm of trust research has shifted from a focus on interpersonal and inter-firm dynamics to include technology, particularly artificial intelligence (AI). Trust is vital in understanding how individuals adopt AI suggestions, especially within systems where AI acts as a decision support tool. This study identifies "trust in AI" as the confidence managers have that these systems will not undermine their interests, indicating a readiness to embrace some level of risk. Drawing from traditional organizational trust theories, trust in AI consists of three key components: ability, benevolence, and integrity. Ability pertains to the AI's proficiency in assisting with complex tasks, benevolence relates to its capacity to enhance human decision-making and the overall welfare of the organization, while integrity involves criteria such as consistency, transparency, and adherence to established guidelines. Here, the decision-maker is characterized as a tech-savvy senior manager, and the AI system serves as the object of trust.
In the study, human-AI collaborative innovation is understood as a series of information processing stages, including gathering, interpreting, verifying, synthesizing, and making decisions. This progression leads to two distinct mechanisms illustrating the relationship between trust in AI and advancements in digital innovation, which follows an inverted U-shaped curve.
The “technology-driven” mechanism suggests that trust in AI enhances efficiency throughout the information processing stages, thereby accelerating digital innovation. The success of digital innovation largely hinges on how well information processing aligns with environmental requirements. When managers trust AI, it can automate routine processes, leading to improved efficiency and more precise decision-making. AI excels in managing vast datasets quickly and affordably, surpassing human cognitive limitations. Consequently, trust reduces barriers to adoption and cognitive costs, enabling actionable insights that drive targeted product iterations and create a streamlined decision-making loop. Beyond simple automation, trust in AI enhances human capabilities, allowing for knowledge augmentation. AI harnesses probabilistic reasoning to suggest ideas, predict trends, and unveil hidden correlations, thus promoting a complementary relationship between human and AI inputs, emphasizing collaboration over replacement. Such synergy is indicative of prior research linking human-AI partnerships to improved performance.
Conversely, the "wicked curse" mechanism reveals that over-reliance on AI can hinder a system’s resilience and error correction abilities, resulting in increasing losses over time. High levels of trust can lead to automation bias, which makes users less critical of AI outputs and hinders the evaluation of information validity and associated ethical risks. This can cause inaccuracies such as algorithmic bias and data distortion to go unnoticed, derailing AI-driven innovations from their intended goals. Additionally, excessive trust may prompt managers to relinquish critical evaluative roles, fostering a fragile decision-making process. Over time, this could diminish their ability to independently assess information, adapt, and make informed decisions in complex contexts. Excessive reliance on AI can also impede a manager's engagement in the essential learning and skill-building necessary for informed decision-making. This prolonged dependence may create cognitive rigidity, diminishing professional judgment, and the ability to think critically, ultimately eroding deep information processing capabilities. Given that human intuition and contextual understanding are crucial for diverse information acquisition, over-trust in rigid AI algorithms may limit the inflow of varied and unconventional inputs, undermining innovation outcomes and competitive edge.
The net effect of these mechanisms is that while the “technology-driven” channel generates increasing benefits with greater trust in AI, the “wicked curse” mechanism invokes mounting risks that escalate rapidly. Thus, the overall relationship between trust in AI and digital innovation manifests as an inverted U-shape. There exists an optimal trust level where the benefits of trusting AI align with its costs, beyond which the challenges imposed by the “wicked curse” overshadow the advantages of the “technology-driven” pathway. Within this ideal range, a harmonious blend of human creativity and AI capabilities can achieve maximum innovation effectiveness. Before reaching this inflection point, the overall impact of trust in AI on innovation climbs as the positive benefits of the “technology-driven” mechanism dominate. After surpassing the threshold, negative aspects from the “wicked curse” installation begin to prevail. Hence, we propose the hypothesis that the connection between trust in AI and digital innovation is characteristically an inverted U-shape.
Further, the study explores how intellectual capital—comprising human, structural, and relational dimensions—influences this relationship. Strong human capital enhances the benefits of the “technology-driven” mechanism while curbing the downsides of the “wicked curse” mechanism by ensuring better information interpretation and scrutiny. High levels of human skill and knowledge allow firms to leverage AI effectively, fostering breakthrough innovations and enabling more rigorous evaluations of AI outputs.
Structural capital also plays a crucial role by embedding risk management protocols, prompting standardized verification processes when trust exceeds healthy boundaries. This safeguards against cognitive inertia and ensures that the dynamic assessment of AI outputs remains intact.
Relational capital enhances the “technology-driven” mechanism by providing access to a diverse pool of information from external sources, which may otherwise be overlooked. This facilitates a broader scope for innovation and acts as a buffer against the narrowing effect wrought by over-reliance on AI.
In conclusion, the interplay of these aspects collectively mitigates the escalating risks associated with the “wicked curse” mechanism and amplifies the benefits of the “technology-driven” approach. The research posits that higher intellectual capital creates a flatter curve in the inverted U-shaped relationship between trust in AI and digital innovation while pushing the turning point where costs begin to outweigh benefits further to the right.
Task complexity also affects this relationship. The challenging nature of innovation tasks may enhance the “technology-driven” mechanism, as AI can assist in processing intricate information and coordinating diverse resources that exceed human cognitive capacity. However, high complexity can intensify the challenges linked to the “wicked curse” mechanism, as difficulties in real-time evaluation of AI outputs can emerge.
Thus, the proposed hypotheses indicate that task complexity raises the marginal benefits from the “technology-driven” aspect while amplifying the costs associated with the “wicked curse” mechanism. This dynamic extends the effective range of trust in AI and shifts the turning point to the right.
Lastly, data quality is a fundamental concept that encompasses attributes like accuracy and reliability. High-quality data is pivotal in facilitating the “technology-driven” mechanism, as AI's effectiveness hinges on the quality of its input. Conversely, poor data quality can heighten risks associated with the “wicked curse,” necessitating a careful balance to optimize the impact of trust in AI on digital innovation. Thus, data quality likely flattens the inverted U-shaped relationship while delaying the point at which the costs override the benefits.
