<p>Addiction involves rapidly fluctuating affective and value-based decision-making processes that undermine cessation efforts, yet capturing these dynamics at clinically meaningful timescales remains challenging. Standard cognitive decision-making tasks are time-intensive and require many trials to achieve reliable parameter estimates, limiting their use longitudinally and in real-world settings. Here, we developed a rapid, smartphone-based framework that integrates ecological momentary assessment (EMA) with adaptive design optimization (ADO), a Bayesian method that enables reliable estimation of computational decision-making parameters from as few as 20-30 trials per task. We tested this framework in <i>N</i> = 79 individuals undergoing a 5-6-week smoking cessation program, who completed daily ADO-based delay discounting and risk/ambiguity tasks alongside EMA surveys assessing smoking behavior, craving, stress, mood, anxiety, and medication adherence. At the day-to-day level, elevated craving, depressive symptoms, and ambiguity tolerance predicted increased smoking the following day, whereas lower discounting rates and reduced craving and stress predicted cessation success at treatment completion. For the latter, models using data from the first week achieved meaningful predictive performance (mean AUC = 0.76), approaching the upper-bound performance observed in models incorporating both task and survey data from the full study period (mean AUC = 0.83). Together, these findings demonstrate that rapid, low-burden, ADO-based delivery of decision-making tasks via EMA can capture clinically relevant, dynamic vulnerability states during smoking cessation treatment. This methodology offers a promising approach for identifying cognitive markers that may facilitate or inhibit cessation success and informing personalized, time-sensitive intervention strategies for nicotine addiction and related psychiatric disorders.</p>