Session: 16-01-01: Government Agency Student Poster Competition

Paper Number: 149672

149672 - From Trailhead to Summit: Using Nlp to Analyze Thru-Hikers’ Wilderness Experiences 

In this study, we apply Natural Language Processing (NLP) as a language analysis approach to examine changes in long-distance hikers’ wilderness experiences. Using the Appalachian Trail, the longest hiking footpath in the world, as our research setting and textual data from Trail Journals, a blog for hiking journal writers, we developed a validated approach to examine changes in hikers’ wilderness experiences over the 6–8-month duration of their 2000-mile journey. We analyzed approximately 150,000 entries from the Trail Journal blog spanning from 2016 to 2021, employing both Borrie and Roggenbuck’s (2001) wilderness experience scale and Kaplan and Talbot’s (1986) wilderness perceptions to measure changes through textual data.

This research is developed as a two-phase study. In the first phase, we replicated Borrie and Roggenbuck’s (2001) study to examine whether the wilderness experience can be measured using a computational approach. First, we created a human-validated lexicon for training the machine learning model on four wilderness experience modes, including the constructs of oneness/primitiveness/humility, timelessness, solitude, and care. Second, we used a supervised machine learning approach to train the NLP model on these lexicon constructs. Third, recognizing that wilderness experiences are dynamic and multiphasic, we divided the textual data into three main phases: entry, immersion, and exit. Fourth, we extracted the different wilderness experience constructs across all three phases using the trained NLP model. Finally, we examined the frequency of these constructs in the different phases.

In the second phase, we followed the same steps as in the first phase but used a developed integrated wilderness scale, combining constructs from both Borrie and Roggenbuck’s study and Kaplan and Talbot’s study. This integrated scale includes constructs such as oneness, awe and wonder, simplicity, solitude, sense of accomplishment, loneliness, tiredness, care for nature, and restoration. This scale provides a more comprehensive view of the different psychological constructs, including negative emotions such as loneliness and tiredness, as well as possible motives such as feelings of accomplishment.

We hypothesized that hikers would experience heightened feelings of oneness/primitiveness/humility at the exit stage, while feelings of care would be prominent during the immersion phase. We also hypothesized that some constructs, such as timelessness, might be challenging to detect using a computational approach. Using the integrative scale might provide more insights into the negative emotions and motivations experienced during the hike. This study provides insights into the changes in emotional states in specific settings over time using language analysis approaches. Furthermore, it introduces an innovative methodological approach that uses online digital records and integrates machine learning techniques with traditional psychological research methods to enhance our understanding of individuals' experiences in wilderness environments. In doing so, we consider the potential applications and limitations of NLP for understanding person-environment relationships using narrative/textual data.

Presenting Author: Norhan Abdelgawad Virginia Tech University

Presenting Author Biography: Norhan Abdelgawad is a PhD student in Urban Affairs and Planning in the School of Public and International Affairs at Virginia Tech. She has an Urban Planning Analytics certificate from Virginia Tech. Her research interests lie in the intersection of art, urban planning and design, environmental psychology, and data analytics. Her dissertation is focused on leveraging crowdsourced and open-source data to evaluate spatial justice in the allocation of cultural resources in Los Angeles. Norhan has a Bachelor of Science in Architecture Engineering from The American University in Cairo and a Master of Science in Architectural Computation from University College London, where she specialized in data and computational design. Norhan has worked as a Landscape Architect and Urban Designer for six years on major mega projects in the Middle East and North Africa region.

Authors:

Norhan Abdelgawad Virginia Tech University
Morva Saaty Virginia Tech University
Jaitun V Patel Virginia Tech University
Kris Wernstedt Virginia Tech University
Scott Mccrickard Virginia Tech University
Shalini Misra Virginia Tech University