In the evolving landscape of building systems and human-building interaction (HBI), the complexity of building interfaces has significantly increased, posing both challenges and opportunities for enhancing energy consumption, indoor environmental quality (IEQ), and building services. This dissertation, titled "User-centered Evaluation of Multi-modal Building Interfaces," delves into the realm of HBI by focusing on the user's experience and perception of multimodal building control interfaces, particularly the various visual modalities of Connected Thermostats (CTs). This body of work aims to support CTs' ongoing adoption, expansion, and performance through a user-centered perspective. The research is motivated by the observation that the design process in the current building industry often overlooks a human-centered approach, leading to a disconnection between occupants' needs and building interface design. This misalignment not only results in user dissatisfaction but also leads to a missed opportunity in leveraging smart building technologies to enhance building performance for achieving climate change mitigation goals. This research attempts to address the main identified gaps in the literature and AEC industry concerning 1) human interaction and perception of multimodal CT interfaces,2) the scarcity of knowledge in the field of human-computer-building interaction (HCBI) regarding the user study methods, 3) the exiting highly non-standard practices in the design of building interfaces. This research highlights 1) the necessity of a multimodal interaction approach, 2) robust mixed-methods User Experience (UX) summative evaluation studies, and 3) the need for standardization in HCBI. This body of work is grounded in the Technology Acceptance Model (TAM) and Human Information Processing (HIP) theories, aiming to foster the adoption of connected building controls with a special focus on usability by suggesting best practices in design and research. The methodology comprised three-step mixed-methods summative evaluation studies designed using a funnel approach to answer the general question: "How do users interact with connected thermostats, and how do these interactions inform our understanding of human-building interaction?": 1) The first and broadest study leveraged texting mining big data of user reviews to identify the general themes and patterns that affect the UX and acceptance of CTs. 2) The second study employed mixed-methods lab experiments to further focus on usability, being recognized as the most determining factor in the adoption of CTs in the first study. This study investigated human interaction with three of the most prevalent modalities of CTs: the Fixed Visual Display (FVD), the phone app, and the web portal. 3) The third study investigated human interaction with a specific visual aspect of UI of FVD and phone app modalities, the interface icons, with the goal of providing some data-driven guidelines for their standardization. Throughout the three studies, the dissertation employed and evaluated some novel and established HCI summative user evaluation methods, including a grounded theory approach for text mining and analyzing user-generated content, eye-tracking think-aloud protocol and contextual inquiry, A/B testing and NASA TLX and SUS surveys to evaluate UX, usability and mental workload. The dissertation outlined three discrete contributions: 1) It bridged some of the well-established UX research methods into HCBI and highlighted the potential of knowledge in the HCI field, 2) Provided guidance for human-centered design of multimodal building interfaces through identifying the main strengths, weaknesses, opportunities, and threats in UX of CTs, 3) Informed the standardization of UI of multimodal building interfaces. / Doctor of Philosophy / Modern building systems are becoming increasingly complex to enhance energy savings and improve occupant convenience. Traditional buildings are typically controlled by wall-mounted devices that manage lighting, temperature, and security. However, in smart buildings, occupants now have new ways to interact with these systems, including remote control options like smartphone apps. This research explores how people interact with and perceive different types of building controls. Specifically, it compares user experiences with two main types of controls: fixed controls, such as wall-mounted thermostats, and mobile controls, like smartphone apps. The goal is to improve the design of these controls to make them easier to use, widely adopted, and more effective in promoting energy efficiency and comfort. Among building controls, smart residential thermostats are particularly important. How they are designed and how people use them can significantly impact energy use in homes. However, advanced thermostats, especially those with programming features, are often seen as difficult to use for building occupants. Designers sometimes overlook the needs of users due to agile design processes or a limited understanding of user-centered design. These complex interfaces can lead to user frustration and missed opportunities to improve building performance. This research addresses these challenges by studying how people actually interact with different smart thermostat controls, including wall-mounted devices, phone apps, and web portals, emphasizing the importance of designing with user needs in mind. This study identifies the common issues users face with the dashboard of these devices, proposes solutions, and offers best practices for creating user-friendly building controls. This study makes three key contributions: First, it introduces well-known methods from other fields for designing and evaluating technology and applies these methods to evaluate the user experience of building controls. Methods include text analysis of user reviews, surveys, eye-tracking, and user performance measurements, some of which are relatively new in the context of human-building interaction. Additionally, the study provides guidance for future researchers in developing a user-centered framework for building interfaces by comparing these methods. Finally, it helps designers by identifying critical usability problems in building interfaces and highlighting opportunities for improvement.