In recent years, we have witnessed an improvement in human happiness. Since the emergence of online assessment and deep learning, computing happiness models to study the contribution of certain factors has become more efficient for ecological momentary assessment (EMA) compared to traditional regression-based methods in social science. In this field, the insights and findings of sociologists are expected to be scientific, and comprehensive enough to deal with almost practical situations. However, the factor explanation by deep learning models somewhat conflicts with these findings. This restricts the application of human-centered artificial intelligence because the results may not be the right ones for the right reasons. To address this issue, we propose a happiness computing solution whose explanation aligns with a set of sociological insights that reflect the core findings cherished by the sociologist community. Two non-trivial research questions are raised: what should the happiness computing models be aligned with? and how to align models with a given goal? Specifically, we summarize the sociological insights, with primary and secondary factor relations as a goal for aligning bytes with bliss, which will guide happiness computing models to find the keys in this ocean of factors. Experimental results using two online datasets and mapping studies with social science literature demonstrate that the computing models align with sociological insights. Our study overall provides more trustworthy happiness factors for assisting human decisions.