We aim to investigate here whether we can develop a reinforcement learning game which provides game-based feedback to encourage positive actions (behaviors) both inside and outside of the game. Does providing positive reward when participants make decisions which are associated with value-based actions (like those in BA) result in different game decisions? We propose that it will increase positive actions in the game. And, secondly, how does it affect short-term behavior (in one week)? We propose that it will increase pro-health activities and may reduce depressive symptoms.

We know that behavior influences mood -- our best interventions to improve mood rely upon the relationship between these. Treatments like this are thought to work in part by helping individuals to increase value-derived behaviors; participants are given guidance which results in an increase of positive behaviors and a decrease of coping behaviors that don't help -- that is, their health-seeking behavior is reinforced while behaviors that diminish health are reduced. In past work, we showed that a text-based game could be used to explore what sort of decisions people would make in certain environments. That game showed associations between in-game behaviors and real-life depressive symptoms and actions. Such work focuses on low-level symptoms of depression -- increasingly common, especially after the onset of the covid-19 pandemic. We aim to investigate here whether we can develop a reinforcement learning game which provides game-based feedback to encourage positive actions (behaviors) both inside and outside of the game.

Thus, the experiment described below and proposed in this application would test the role of positive rewards (positive-feedback) in a dichotomous-choice game, compared to neutral (no-feedback).

Inclusion Criteria:

• 18-34 years old
• Fluent in English
• Based in the United States

Exclusion Criteria:

• None

Data will be shared on the Open Science Framework, https://osf.io

Data will be available on or before a manuscript is submitted for publication and will be shared indefinitely.

Open access. Anyone will be able to access de-identified data and code.

The intervention occurs entirely online. All assessment occurs via survey software. Participants do not interact with other participants and are not fully informed of the nature of the intervention and thus do not know if they are in an active condition.