The drive for improved building energy performance has led to increased automation levels in building control systems, challenging the balance between energy efficiency and comfort. This has often resulted in decreased user satisfaction with indoor environments for two main reasons, occupants are unable to adjust conditions to their preferences, resulting in discomfort, and the lack of control causes psychological dissatisfaction. Since people spend approximately 90% of their time indoors, shifting from technology-centric to human-centric control (HCC) systems in buildings is crucial. HCC allows the integration of user preferences into the building control system for indoor environmental control. This approach not only improves occupant comfort but also enables occupants to change their energy consumption behaviour towards more efficient and sustainable behaviour. This is particularly beneficial for demand-response actions in imbalanced grids, where load fluctuates during peak and off-peak hours, and there is a higher participation of renewable energy sources, sometimes resulting in a surplus of generated electricity. As buildings account for about 75% of electricity consumption in developed countries, with 80% of that occurring during peak hours, engaging users in grid flexibility actions through advanced control systems like HCC shows significant potential for supporting grid efficiency and stability. However, this integration of user preferences into controllers requires occupants to interact with the system and share data or feedback on their perceived comfort or to engage in grid flexibility actions, which can be inconvenient for users. Data-driven personalized comfort models (PCMs) offer an optimal solution by learning occupant preferences (e.g., if one wants warmer or cooler, dimmer or brighter, comfortable or not) and representing their comfort levels without requiring continuous interaction. Nonetheless, this still demands some effort from users until sufficient data is collected to learn their preferences. In HCC research, seamless user interaction and continuous data sharing are often assumed, leading to a lack of user-provided data when deployed in real scenarios. This thesis explores user perspectives on interacting with HCC, using a survey-based study to examine their opinions and perceptions and their willingness to share data or feedback for indoor environmental control and grid flexibility interaction. It developed a method to create PCMs for multiple comfort aspects and multiple occupants during two field experiments, including indoor air quality, thermal, and visual comfort PCMs. During the second field experiment, the impact of implementing HCC in an actual building is tested, providing important insights from the real experiences of 24 building occupants interacting with HCC. Additionally, through simulations, the study investigates how user-building interaction facilitated by HCC coupled with advanced controllers like Model Predictive Control (MPC) can be leveraged for grid flexibility and the impact of HCC on grid flexibility, energy consumption, costs, and comfort. The research methods include a survey-based study with almost 1 000 respondents, a behavioural science theory, two field experiments, and simulations. This emphasizes the interdisciplinary nature of this thesis, which bridges technical and social sciences. The survey was conducted in Zagreb, Croatia, with occupants from six educational buildings consisting of a diverse target group. The first field experiment involved four participants in a faculty office over three weeks, gathering data to develop PCMs. The second experiment took place in a high school building (RCK Ruder Boskovic) in Zagreb, Croatia, with 24 participants over two weeks, aiming to test the findings from the first experiment in real scenarios and integrate user preferences into the open code building control system. A mixed-method approach consisting of mini-surveys and semi-structured interviews evaluated the impact of implementing HCC on user satisfaction during the school field experiment. Lastly, simulations compared the effects of integrating user preferences into conventional PID control and advanced systems like MPC. More importantly, the impact of combining HCC with MPC and engaging users in demand response or grid flexibility actions was analyzed and quantified. The survey-based research revealed that 75.7% of nearly 1 000 occupants want access to building controls, but only 55.6% are willing to interact with HCC by sharing data. Indoor air quality was the most important comfort aspect for 85% of occupants, followed by thermal comfort (84%) and visual comfort (74%). A behavioral science model with 64% prediction accuracy was developed to predict occupants’ willingness to share data with HCC. This study also elicited that the key factors influencing data sharing with HCC, ranked by impact, are beliefs about its usability and benefits, ease of use, social influence, and privacy security. The survey also indicated that 66% of respondents prefer using smartphones as interfaces and primarily interact with controls only when discomfort arises. Field experiments, designed based on these preferences, collected data from actual building occupants in natural environments. The results from these experiments show that tree-based models, like Decision Trees and Random Forest, are the best-performing machine learning models for developing PCMs, depending on the varying levels of occupant engagement in providing data. The school field experiment indicated that integrating user preferences into the control system increased user satisfaction with the indoor environment by 16.7%. Regarding grid flexibility, 59.5% of survey respondents were willing to engage in grid flexibility or demand response actions. About 47.7% preferred interacting for 1 – 2 hours daily, with varying expectations for financial incentives, even if it meant adjusting indoor temperatures to unfavourable conditions during grid overloads to reduce energy usage. Simulations combining HCC and MPC with grid flexibility actions while prioritizing occupant preferences achieved 44.6% energy savings, 60.6% load shifting, and 59.4% cost savings during peak hours compared to the baseline PID scenario with standard setpoints. Comfort improved by 38% compared to the baseline, yet the improvement was lower compared to scenarios that primarily focus on occupant preferences without engaging in grid flexibility, which is expected and assumed to be agreed upon by occupants. These findings are valuable for HCC designers, practitioners, and stakeholders aiming to create control systems prioritizing occupants and their preferences. Additionally, insights on leveraging user-building interaction through HCC for grid flexibility are crucial for the future role of buildings in smart and sustainable grids.