HUMAN-IN-THE-LOOP VS. LIGHTS-OUT AUTOMATION

As the global financial landscape is transitioning from a traditional ecosystem to digitally based transactions rapidly, AI models for scam detection have turned into an indispensable pillar for proactive defence. The opaqueness of automated decision-making has however caused the Human Trust conundrum that cannot be addressed by algorithmic models alone. To find out the levels of trust among the users, this paper involves comparison of consumer trust in human-in-the-loop systems versus that of completely independent systems.

The mean trust score of AI systems of the survey was calculated using inferential statistics and a 5-point Likert scale, the measure of trust, and provided a mean of 3.04 and a standard deviation of 0.30, compared to the average score of a human-led systems at 2.97 and a standard deviation of 0.32. The t-test of paired samples gave a value of 0.5984, that is, there is no considerable difference between the levels of trust between the two interventions. Moreover, ANOVA tests confirmed that trust level do not differ based on age group. Chi-square provided a p-value of 0.0529, that is, there is a strong marginal tendency preferring age- based preference, although 53.45% of the participants chose instantaneous AI blocking as opposed to human verification (46.55%). The result of these findings is a situation where trust is balanced, where users perceive that the respective systems are equally reliable, despite the possibility that functional preferences for speed may be directed towards a gradual shift to automated solutions.