As data grows complex, making sense of complex data for problem-solving by teams is becoming a challenge. Previous research suggests that challenging data problems can only be solved by leveraging human cognition, in combination with computational advances. However, this potential remains untapped, as collaborative sensemaking is fraught with significant multiple socio-cognitive challenges of information sharing and analysis. Lack of human centered design and evaluation approach to develop information sharing and problem-solving tools have resulted in little empirical knowledge about these challenges and potential design solutions to overcome these challenges. This dissertation offers a human centered design approach to iteratively design and evaluate collaborative sensemaking tools for a problem-solving task in the crime-solving domain. Crime-solving domain offers a life-critical germane ground to design for known human challenges that continue to recur. For every challenge, I designed and deployed a tool, and evaluated its effectiveness in a laboratory experiment where participants used my tools to solve a crime problem collaboratively and reported on task-performance and collaboration experience. Tools deployed in each iteration benefitted from the objective results, self-reported perception, user-log analysis, video-analysis, and qualitative feedback, from the previous iteration. First, I designed SAVANT as a modular tool to highlight that data analytic tools perform better when customized for simplicity to enable different sensemaking tasks at hand, as opposed to offering all the complex features always. Next, SAVANT was modified based on lab experiment to solve the social challenge of inefficient explicit sharing of information among crime analysts. Collaborative version of SAVANT offers implicit sharing of notes and insights as an alternative between remotely collaborating data analysts. Finally, I invented Sensemaking Translucence, a design metaphor used to overcome the cognitive challenge of biased decisionmaking through implicit visualization of decision-process artifacts, deployed in REFLECTIVA. By leveraging intermediate data analytic artifacts, including notes, insights, and communication, to drive visualizations in SAVANT and REFLECTIVA, my findings would benefit the design of web based data analytic tools. Future research directions and design implications deriving from these findings are also outlined at the end of this dissertation.