Human Choice Predicted by Obtained Reinforcers, Not by Reinforcement Predictors

Abstract

Macphail (1985) proposed that “intelligence” should not vary across vertebrate species when contextual variables are accounted for. Focusing on research involving choice behavior, the propensity for choosing an option that produces stimuli that predict the presence or absence of reinforcement but that also results in less food over time can be examined. This choice preference has been found multiple times in pigeons ( Stagner and Zentall, 2010 ; Zentall and Stagner, 2011 ; Laude et al., 2014 ) and has been likened to gambling behavior demonstrated by humans ( Zentall, 2014 , 2016 ). The present experiments used a similarly structured task to examine adult human preferences for reinforcement predictors and compared findings to choice behavior demonstrated by children ( Lalli et al., 2000 ), monkeys ( Smith et al., 2017 ; Smith and Beran, 2020 ), dogs ( Jackson et al., 2020 ), rats ( Chow et al., 2017 ; Cunningham and Shahan, 2019 ; Jackson et al., 2020 ), and pigeons ( Roper and Zentall, 1999 ; Stagner and Zentall, 2010 ). In Experiment 1, adult human participants showed no preference for reinforcement predictors. Results from Experiment 2 suggest that not only were reinforcement predictors not preferred, but that perhaps reinforcement predictors had no effect at all on choice behavior. Results from Experiments 1 and 2 were further assessed using a generalized matching equation, the findings from which support that adult human choice behavior in the present research was largely determined by reinforcement history. Overall, the present results obtained from human adult participants are different than those found from pigeons in particular, suggesting that further examination of Macphail (1985) hypothesis is warranted.

Suboptimal choice: A review and quantification of the signal for good news (SiGN) model.

“…They found strong preference for an optimal alternative that always provided three points over a suboptimal alternative that provided 10 points 20% of the time, and this preference occurred both when the suboptimal alternative was signaled and when it was unsignaled. Stagner et al (2020) also tested adult humans and found results similar to McDevitt et al (2019). In sum, in standard suboptimal choice procedures, adult humans seem less likely than pigeons to choose suboptimal alternatives, but more studies are needed.…”

The Effect of Non-Instrumental Information on Reward Learning

“…Given that scenarios in which outcomes are learned from experiencerather than descriptionare ubiquitous in everyday settings, examining human choices preferences in paradigms more closely resembling those in the animal literature (i.e., outcomes learned via experience where one option is suboptimal) remains pertinent. Although this question has been addressed in some recent work (Molet et al, 2012;McDevitt et al, 2019;Stagner et al, 2020), studies have yielded inconclusive results and used very small sample sizes (ns < 20) 1 .…”