Skip to main content
Open Access Publications from the University of California

UC Irvine

UC Irvine Electronic Theses and Dissertations bannerUC Irvine

Improving Efficacy of Support Groups in Online Environments

Creative Commons 'BY' version 4.0 license

In this research I evaluated two strategies for improving performance of online support groups. The first strategy was to use a buddy system in online support groups. This system involves pairing demographically similar members of a group to serve as each other’s buddies. Analyses of real online support groups indicated that members whose only difference with other members is in having a buddy who is more active rather than less active in the group are more engaged in the group and have a higher chance of attaining their goal for which they joined the group. Analysis also showed that members help their buddies more than they help other group members. Results are robust across four measures of tie strength, including contact frequency, reciprocity, and two measures of contact length. Overall, results suggest that managers can use the buddy system in online support groups as an effective method to drive group engagement and increase the support provided and goal attainment.

The second strategy was to add a chatbot to the online support groups to provide members with additional informational and motivational support. The design of the chatbot which can respond to members’ messages based on their content is presented here. The chatbot has a natural language understanding component which can identify intent of messages out of 26 possible intents related to smoking cessation which is the main context of the online support groups studied here. The chatbot is supposed to respond to messages with 25 intents called triggering intents and ignore 1 intent which is non-triggering intent. Triggering intents account for less than half of the total messages. The bot is intentionally designed to have higher precision than recall for triggering intents. Precision for triggering intents is the probability of identifying the intent correctly if the message is identified as a triggering intent. Recall for triggering intents is the percentage of triggering messages which are identified with correct intent. The chatbot has the precision of 81% and the recall of 44% for triggering intents. Separate randomized control trial experiments are required to evaluate the overall performance of the chatbot in online support groups.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View