Building trust in the age of human-machine interaction: insights, challenges, and future directions

Abstract

Trust is a foundation for human relationships, facilitating cooperation, collaboration, and social solidarity (Kramer, 1999). Trust in human relationships is generally based on factors like dependability, competence, generosity, and sincerity (Mayer et al., 1995; Lewicki & Bunker, 1996). Social norms, emotional intelligence, and the power of forecasting others' behaviors help create shared knowledge and mutual respect (Coleman, 1990; Rotter, 1980).As technology more and more becomes incorporated into everyday life, especially by means of artificial intelligence (AI) and robotics, the concept of trust has shifted paradigmatically (Lankton et al., 2015). In Human-Robot Interaction (HRI), trust does not derive from emotional familiarity or social intuition but rather from properties of the system itself, such as functionality, transparency, and predictability (Hancock et al., 2011). This invokes basic questions: Can humans ever trust machines? If they can, how is that trust established, sustained, or dissolved?There is growing evidence that humans can work with robots in various situations, such as search-and-rescue missions, education, and healthcare (Breazeal, 2003; Chen & Barnes, 2014; Nagpal et al., 2024; Nandanwar & Dutt, 2023). For instance, latest research utilizing Proximal Policy Optimization (PPO) and Generative Adversarial Imitation Learning (GAIL) identify that robots have the ability to excel over human peers in difficult search-and-retrieve tasks in a situation where trust is calibrated (Kapoor et al., 2024a; 2024b). In the same vein, emotionally responsive robots have demonstrated potential in improving language learning achievement in school children (Nagpal et al., 2024), whereas affective conversational agents assist in stress and anxiety reduction in patients (Nandanwar & Dutt, 2023).But embedding AI systems within fields such as autonomous driving, military action, and healthcare introduces novel trust challenges. These are the opacity of decision-making by algorithms, variable levels of autonomy, and cultural compatibility clashes in user expectations (Chen et al., 2018; Goodall, 2014; Schaefer et al., 2016). Even when AI is reliable, a lack of explainability will undermine user trust. Consequently, Explainable AI (XAI) is essential in closing the cognitive and affective space between humans and machines (Arrieta et al., 2020).However, trust in HRI is not always built. It differs by cultural environment, personality type, and task context. Although tremendous strides have been made in the modeling of trust as a system performance function, current models tend to overlook dynamic, emotional, and socio-cultural aspects (Eiband et al., 2018; Hoff & Bashir, 2015).This opinion paper contributes to the discussion by comparing the building blocks of trust in human-human and human-robot interaction. It presents the Trust-Affordance Adaptation Model (TAAM)—a theoretical framework that aligns trust-building tactics with domain requirements. We contend that emotional investment and functional openness need to be traded off depending on context, and we propose the incorporation of psycho-social cues, like biosensor information, into trust modeling. Through a synthesis of current literature and findings of recent empirical research, the paper provides a guide for developing reliable AI systems that are emotionally engaged, culturally adaptable, and context-sensitive.2. Trust in Human-Human InteractionSeveral basic disciplines, such as organizational behavior, psychology, and sociology, have thoroughly researched the interpersonal trust phenomena (Lewis & Weigert 1985; Rotter 1980). Based on Figure 1, it appears that there are many basic factors to consider that facilitate or sustain relationship trust. Establishing and maintaining trust is particularly difficult in business settings. Dependability is the primary trait, especially in firms where cooperation and production rely on one another (Mayer et al., 1995; Dirks & Fe