This thesis explores the use of Augmented Reality (AR) in two challenging scenarios -- autism therapy, and network-constrained, remote collaboration. In the first half of this thesis, we highlight the design and implementation of CustomAR and Augmented Playgrounds -- two applications that leverage the visualization and contextual abilities of AR in autistic contexts. The former allows therapists and educational professionals to augment pictures and visual supports with virtual content for use in autism-based learning exercises. The latter introduces interactive AR social stories that therapists could use as potential tools to present social situations and teach social skills to autistic students in real-world environments, such as playgrounds. We use participatory design practices and exploratory interviews to understand user needs, especially that of customization, and develop appropriate features for each application. Our user studies focus on evaluating the applications with therapists and educational professionals, and gauging whether the customization options are sufficient to allow them to adapt the content of the AR experiences to meet the varied needs and interests of their autistic students, and what additional options are necessary. We also highlight the practical challenges users may face when trying to use AR in therapy/learning settings, or in outdoor physical environments, such as playgrounds. Similarly, remote collaboration is an important use-case for AR as it can enable geographically distant users to collaborate using shared video feeds or interactive, content-rich 3D holograms. However, network impairments are inevitable in today's best-effort internet and could adversely affect user experience. The latter part of this thesis, therefore, focuses on understanding how network outages affect user experience in AR-based remote collaboration. Through an empirical, in-lab user study, we highlight the negative effects of outage, and the strategies users adopt to deal with them, such as batching instructions or changing spatial referencing styles. We also highlight how giving users explicit information about poor network conditions helps them reduce the extra effort they may need to put into the collaboration. Towards the end, we present some design implications for future remote-collaborative AR applications and discuss the implications of our findings for developing regions, where network impairments are more common. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://bibliotheek.ehb.be:2222/en-US/products/dissertations/individuals.shtml.]