Trust and Cooperation

Abstract

We AI researchers are concerned about the potential impact of artificially intelligent systems on humanity. In the first half of this essay, I argue that ethics is an evolved body of cultural knowledge that (among other things) encourages individual behavior that promotes the welfare of the society (which in turn promotes the welfare of its individual members). The causal paths involved suggest that trust and cooperation play key roles in this process. In the second half of the essay, I consider whether the key role of trust exposes our society to existential threats. This possibility arises because decision-making agents (humans, AIs, and others) necessarily rely on simplified models to cope with the unbounded complexity of our physical and social world. By selecting actions to maximize a utility measure, a well-formulated game theory model can be a powerful and valuable tool. However, a poorly-formulated game theory model may be uniquely harmful, in cases where the action it recommends deliberately exploits the vulnerability and violates the trust of cooperative partners. Widespread use of such models can erode the overall levels of trust in the society. Cooperation is reduced, resources are constrained, and there is less ability to meet challenges or take advantage of opportunities. Loss of trust will affect humanity’s ability to respond to existential threats such as climate change.

The importance of need‐altruism and kin‐altruism to blood donor behaviour for black and white people

“…As there are no existing quantitative data on kin-and need-altruism by ethnicity, we based calculations on variation in trust in individuals by ethnicity reported by Ferguson et al 7 Trust in individuals was seen as an appropriate index as it underlies altruism and cooperation generally. 42 The effect size for the comparison across the four ethnic groups (Asian, Black, Mixed, White) reported in Ferguson et al 13 equates to a Cohen's d of 0.4871, indicating that for a comparison across the five groups, 66 people are required per group. Based on Ferguson et al, 7 the effect size comparing a White and overall Ethnic minority sample was a Cohen's d of 0.363, indicating that 120 people per group are required for these comparisons reported in studies 1 and 2.…”

“…Power analysis : Power calculations were conducted to achieve 0.80 power with an α of 0.05 (two‐tailed). As there are no existing quantitative data on kin‐ and need‐altruism by ethnicity, we based calculations on variation in trust in individuals by ethnicity reported by Ferguson et al 7 Trust in individuals was seen as an appropriate index as it underlies altruism and cooperation generally 42 . The effect size for the comparison across the four ethnic groups (Asian, Black, Mixed, White) reported in Ferguson et al 13 equates to a Cohen's d of 0.4871, indicating that for a comparison across the five groups, 66 people are required per group.…”

Can robots be trustworthy?

“…Trust is often seen as a design component for successful social interaction and should thus be enhanced to optimize the utility (Kuipers 2022;Billings et al 2012) as well as the acceptability (Whelan et al 2018; Siau and Wang 2018) of the robot. As a consequence, trust is operationalized into a measurable parameter for evaluating the design and the probability of the human to use the device (Hancock et al 2011), making 'social' an experience that can be constructed and consumed.…”

Technical Basis for Advanced Artificial Intelligence and Machine Learning Adoption in Nuclear Power Plants

“…When reaffirming that AI is in existence for the benefit of people, the European Commission HCAI expert panel has stated that AI should be designed not as an end in itself, but rather, as a promising vehicle that together increases the human potential and societal well-being [30]. To achieve this potential, [35] considers trust and cooperation between human and artificial agents as essential. Trust in AI is a central component of HCAI.…”