Information technologies are essential tools for the representation and communication of human knowledge. However, there are significant disparities in the sources of knowledge currently represented on the Internet. To address this, my research group developed Awaaz.De, a phone-based voice message board allowing small farmers in India to ask, answer and listen to agricultural questions and answers. Awaaz.De has been deployed in Gujarat, India for over four years, and receives hundreds of calls every week. I report on recent results from this deployment, including empirical evidence of impact on farmer decision-making, such as reducing the use of less effective and potentially harmful pesticides. While Awaaz.De illustrates the importance of appropriate user interfaces for representing situated knowledge from underrepresented groups, this knowledge must be translated to structured, quantitative forms for aggregation and policy decision-making. Local Ground is a data collection, mapping and information visualization tool that helps youth develop data skills by making connections between different representations of the same empirical phenomena. Students begin by collecting open-ended qualitative data, in the form of free-hand drawings, pictures and audio interviews. This data is iteratively categorized and tagged, leading to the design of structured data collection instruments for more systematic inquiry. Finally, these various forms of data are combined into narratives that can articulate youth perspectives to a variety of stakeholders. Local Ground has thus far been used to contribute to the planning of a public park, ground-truth civic data about food access, and document air quality issues across the BART transportation system. Within these projects, I explore several themes in my work, including the design of more accessible interaction techniques allowing new populations to author content, the importance of bottom-up data for planning and evaluating development projects, and how we can bridge between diverse knowledge representations using participatory data processing techniques.