Despite recent progress in computer vision, fine-grained interpretation of satellite images remains challenging because of a lack of labeled training data. To overcome this limitation, we construct a novel dataset called WikiSatNet by pairing geo-referenced Wikipedia articles with satellite imagery of their corresponding locations. We then propose two strategies to learn representations of satellite images by predicting properties of the corresponding articles from the images. Leveraging this new multi-modal dataset, we can drastically reduce the quantity of human-annotated labels and time required for downstream tasks. On the recently released fMoW dataset, our pre-training strategies can boost the performance of a model pre-trained on ImageNet by up to 4.5% in F1 score.

Progress on the UN Sustainable Development Goals (SDGs) is hampered by a persistent lack of data regarding key social, environmental, and economic indicators, particularly in developing countries. For example, data on poverty - the first of seventeen SDGs - is both spatially sparse and infrequently collected in Sub-Saharan Africa due to the high cost of surveys. Here we propose a novel method for estimating socioeconomic indicators using open-source, geolocated textual information from Wikipedia articles. We demonstrate that modern NLP techniques can be used to predict community-level asset wealth and education outcomes using nearby geolocated Wikipedia articles. When paired with nightlights satellite imagery, our method outperforms all previously published benchmarks for this prediction task, indicating the potential of Wikipedia to inform both research in the social sciences and future policy decisions.

In this paper we describe the results of a design-driven study of collaborative ideation. Based on preliminary findings that identified a novel digital ideation paradigm we refer to as chainstorming, or online communication brainstorming, two exploratory studies were performed. First, we developed and tested a distributed method of ideation we call cheatstorming, in which previously generated brainstorm ideas are delivered to targeted local contexts in response to a prompt. We then performed a more rigorous case study to examine the cheatstorming method and consider its possible implementation in the context of a distributed online ideation tool. Based on observations from these studies, we conclude with the somewhat provocative suggestion that ideation need not require the generation of new ideas. Rather, we present a model of ideation suggesting that its value has less to do with the generation of novel ideas than the cultural influence exerted by unconventional ideas on the ideating team. Thus brainstorming is more than the pooling of "invented" ideas, it involves the sharing and interpretation of concepts in unintended and (ideally) unanticipated ways.