The investigators will send participants supportive text-messages for a period of 2 months. These text-messages will include tips about behavioral activation and coping skills to deal with worries and stress. The investigators generated a message bank balanced such that 50% of all messages are related to behavioral activation (BA) and 50% messages involve different coping skills. Participants will receive one of these messages per day between 9:00 am and 6:00 pm. Participants will also receive a message asking them to rate their mood on a scale of 1-9 once a day 3 hours after the BA or coping message.
Participants will be randomized to:
1. a uniform random messaging group (micro-randomized trial design).
2. a reinforcement learning group with a learned decision mechanism for the timing and type of text-message. The algorithm learns from previous data (which messages were sent, what was the participants' mood) to maximize an increase in participants mood. No other data are collected from participants' phones.
The investigators will compare the effect of sending text-messages by a random schedule, and text-messaging chosen by the RL algorithm. This allows to both evaluate the effect of the individual intervention components over time within a micro-randomized trial design, and assess the added value of using RL to adapt the messaging scheme.
The investigators hypothesize that:
* Participant will show improvements in depression, anxiety symptoms and mood during the 60 day study.
* The participants in the group receiving reinforcement learning will have a greater improvement in depressive symptoms, anxiety and positive mood during the study than participants in the micro-randomized group.
* The investigators will find differential effects on mood ratings for the two categories of messages
