Shorter Planning Depth and Higher Response Noise During Sequential Decision-Making in Old Age

Steffen, Johannes; Marković, Dimitrije; Glöckner, Franka; Neukam, Philipp T.; Kiebel, Stefan J.; Li, Shu-Chen; Smolka, Michael N.

doi:10.21203/rs.3.rs-2095779/v1

Cited by 1 publication

(5 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For distributions of PD across groups and step condition see Figure 3. Consistent with previous findings (Steffen et al, 2022), we observed neither a significant effect of the noise condition ( t (182)=-1.253, p =.21) nor a significant interaction effect of group x noise ( t (182)=1.086, p =.28).…”

Section: Resultssupporting

confidence: 88%

“…Participants entered their responses via keyboard with key ‘S’ indicating the jump action and the right arrow key indicating the move action. For an exhaustive description of the SAT see Steffen et al (2022).…”

Section: Methodsmentioning

confidence: 99%

“…As an analytical solution for the posteriors of the parameters was intractable, we instead used the stochastic variational inference scheme from the probabilistic programming library Pyro v1.5.2 (Bingham et al, 2019) to approximate the posterior distributions. For more details on the computational model and inference procedure see Steffen et al (2022).…”

Section: Methodsmentioning

confidence: 99%

“…One option to approach these decisions is model-based forward planning, which rests upon knowledge about the consequences and a weighing of their negative and positive values (Dolan and Dayan, 2013; Sutton and Barto, 2018). One obvious challenge of forward planning is the question of how deep one should plan ahead (Steffen et al, 2022). Deeper forward planning leads to more accurate valuations of consequences but is also computationally more costly.…”

Section: Introductionmentioning

confidence: 99%

“…To assess PD in the described sample in comparison to HC, we used our sequential decision-making task, the Space Adventure Task (SAT), as described already by Steffen et al (2022). Participants' choices in this task were modelled with a reinforcement learning agent model with PD as a free parameter that could be inferred with Bayesian Inference (see Methods section).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Forward planning under uncertainty in a population-based alcohol use disorder sample

Fischbach

Gönner

Steffen

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Altered decision-making is a defining component of addiction, but there is little evidence whether these alterations affect multi-step planning in individuals with alcohol use disorder (AUD). We used a recently developed planning task in a cross-sectional approach to test the planning performance of 30 individuals diagnosed with AUD relative to 32 healthy control subjects, both sampled from the general population. To gain insight into the factors underlying behavioral performance, we inferred the parameters of a reinforcement learning agent performing rational planning, using a Bayesian inference scheme. This approach allowed us to differentiate between separate factors determining planning performance, e.g., planning depth, decision noise, and bias for certain action choices. Contrary to our hypothesis, we did not observe reduced planning depth in AUD subjects. Instead, our results show a small effect in the opposite direction: Healthy controls were slightly less efficient in the planning task. Importantly, subjects in the control group allocated less time for planning than AUD subjects, potentially indicative of motivational differences between groups. The group difference in planning depth persisted when controlling for both reaction times and general cognitive performance, albeit at a lesser magnitude. Altogether, our results do not favor the view that mild-to-moderate alcohol use disorder in general-population individuals generally involves impairments in cognitive tasks requiring forward planning across multiple steps.

show abstract

Section: Resultssupporting

confidence: 88%

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations