Normative messaging interventions have proven to be a cost-effective strategy for promoting pro-environmental behaviors. The effectiveness of normative messages is partially determined by how personally relevant the comparison groups are as well as the lag of feedback. Using readily available energy use data has created opportunities to generate highly personalized reference groups based on households’ behavioral patterns. Unfortunately, it is not well understood how data granularity (e.g., minute, hour) affects the performance of behavioral reference group categorization. This is important because different levels of data granularity can produce conflicting results in terms of group similarity and vary in computational time. Therefore, this research aims to evaluate the performance of clustering methods across different levels of temporal granularity of energy use data. A clustering analysis is conducted using one-year of energy use data from 3,000 households in Holland, Michigan. The clustering results show that behavioral reference groups become the most similar when representing households’ energy use behaviors at a six-hour interval. Computationally, less granular data (i.e., six and twelve hours) takes less time than highly granular data which increases exponentially with more households. Considering the enormous scale that normative messaging interventions need to be applied at, using less granular data (six-hour intervals) will permit interveners to maximize the effectiveness of highly personalized normative feedback messages while minimizing computation burdens.
more »
« less
Non-Invasive Behavioral Reference Group Categorization Considering Temporal Granularity and Aggregation Level of Energy Use Data
Within residences, normative messaging interventions have been gaining interest as a cost-effective way to promote energy-saving behaviors. Behavioral reference groups are one important factor in determining the effectiveness of normative messages. More personally relevant and meaningful groups are likely to promote behavior change. Using readily available energy-use profiles in a non-invasive manner permits the creation of highly personalized reference groups. Unfortunately, how data granularity (e.g., minute and hour) and aggregation (e.g., one week and one month) affect the performance of energy profile-based reference group categorization is not well understood. This research evaluates reference group categorization performance across different levels of data granularity and aggregation. We conduct a clustering analysis using one-year of energy use data from 2248 households in Holland, Michigan USA. The clustering analysis reveals that using six-hour intervals results in more personalized energy profile-based reference groups compared to using more granular data (e.g., 15 min). This also minimizes computational burdens. Further, aggregating energy-use data over all days of twelve weeks increases the group similarity compared to less aggregated data (e.g., weekdays of twelve weeks). The proposed categorization framework enables interveners to create personalized and scalable normative feedback messages.
more »
« less
- Award ID(s):
- 1705273
- NSF-PAR ID:
- 10175382
- Date Published:
- Journal Name:
- Energies
- Volume:
- 13
- Issue:
- 14
- ISSN:
- 1996-1073
- Page Range / eLocation ID:
- 3678
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Psychologists hypothesize that the effectiveness of normative messaging interventions increases when individuals have more personal attachment and similarity with reference groups. Using readily available energy consumption data, it is now possible to create highly personalized reference groups based on households’ daily energy use in a non-invasive matter. However, it still remains unclear to what degree individuals perceive behavioral reference groups as a cohesive entity. Therefore, this research investigates how individuals perceive energy profile-based groups relative to more standard geographic proximity-based groups. An online survey is conducted with 1,928 U.S. adults. Individuals do not perceive the profile-based groups as very entitative groups. Also, similarity between energy profile-based group members indirectly affects individuals’ identification with the groups via group entitativity. Lastly, this indirect effect is larger than the direct effect of similarity between group members on group identification. These results imply that a better understanding of what affects group entitativity would allow interveners to create more effective normative feedback messages.more » « less
-
Reporting normative feedback to residential energy consumers has been found effective at reducing residential energy consumption. Upon receiving normative feedback households tend to modify their use to become in line with group norms. The effect of normative messages is partially moderated by how personally relevant normative reference groups are to the individual. Advanced energy metering technologies capture households’ energy use patterns, making it possible to generate highly similar and relevant normative reference groups in a non-invasive manner. Unfortunately, it is not well understood how similar individuals are to other group members. It also remains unknown how much individuals identify with behavioral reference groups. Therefore, this research aims to investigate how households perceive behavioral reference groups used in normative comparisons. Survey questionnaires are collected from 2,008 participants using Amazon Mechanical Turk. It is found that while households’ behaviors are more similar when grouped based on energy use profiles than based on geographic proximity, they identify more closely with proximity-based groups. Also, members’ group identification increases as individuals have higher similarity in energy use behaviors with other group members. This implies that enhancing the identity of profile-based behavioral reference groups will lead to an increase in norm adherence, and in turn reductions in household energy use.more » « less
-
Recommender Systems ( RecSys ) provide suggestions in many decision-making processes. Given that groups of people can perform many real-world activities (e.g., a group of people attending a conference looking for a place to dine), the need for recommendations for groups has increased. A wide range of Group Recommender Systems ( GRecSys ) has been developed to aggregate individual preferences to group preferences. We analyze 175 studies related to GRecSys . Previous works evaluate their systems using different types of groups (sizes and cohesiveness), and most of such works focus on testing their systems using only one type of item, called Experience Goods (EG). As a consequence, it is hard to get consistent conclusions about the performance of GRecSys . We present the aggregation strategies and aggregation functions that GRecSys commonly use to aggregate group members’ preferences. This study experimentally compares the performance (i.e., accuracy, ranking quality, and usefulness) using four metrics (Hit Ratio, Normalize Discounted Cumulative Gain, Diversity, and Coverage) of eight representative RecSys for group recommendations on ephemeral groups. Moreover, we use two different aggregation strategies, 10 different aggregation functions, and two different types of items on two types of datasets (EG and Search Goods (SG)) containing real-life datasets. The results show that the evaluation of GRecSys needs to use both EG and SG types of data, because the different characteristics of datasets lead to different performance. GRecSys using Singular Value Decomposition or Neural Collaborative Filtering methods work better than others. It is observed that the Average aggregation function is the one that produces better results.more » « less
-
This study provides a proof-of-concept for a new method for analyzing intonational form and meaning, demonstrated by analysis of mirative utterances in American English. Here, K-means clustering using measures derived from PoLaR labels (i.e., TCoG) revealed emergent clusters of pitch accents that are suggestive of familiar phonological categories (e.g., MAE_ToBI L+H*). A Random Forest analysis then classified utterance-level meaning based on measures from both smaller granularity (related to individual pitch accents) and larger granularity (related to utterance level meaning), showing >85% correct categorization of exclamative vs filler sentences. This work has implications for how to model mappings between prosody and meaning, especially where existing phonological categories alone don’t identify semantic/pragmatic categories.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.3390/en13143678
