Research shows that people frequently try to search for information with other people. The fact that no user interface for collaborative searching has yet caught fire suggests that the best parts of the design space have yet to be investigated. © 1970-2012 IEEE.

Modern Web services are typically free and open access, often supported by advertising revenue. These attributes, however, leave services vulnerable to many forms of abuse, including sending spam via Web-based email accounts, inflating page rank scores by spamming backlinks on blogs, etc. However, many of these schemes are nontrivial to execute, requiring technical expertise and access to ancillary resources (e.g. IP diversity, telephone numbers, etc.). Thus, many scammers prefer to offload the execution of their abuse schemes onto hired labor. This desire to minimize effort has created a demand for workers to carry out malicious tasks. Meanwhile, various online labor marketplaces have emerged that connect employers with cheap, human workers. Abusers have turned to online freelancing sites to find workers willing to carry out numerous schemes. Outsourcing is an attractive option for entrepreneurial scammers, as the workers are typically cheap, technically adept, and exist in vast numbers. In this dissertation, we investigate how outsourcing impacts the security of Web services; no longer must service providers be wary of automated tools, they must now contend with inexpensive human labor willing to do any menial task. In the first part of the dissertation, we characterize the role of freelance labor in Web service abuse, analyzing over seven years of data from the popular crowdsourcing site Freelancer.com, as well data from our own active job solicitations. We identify the largest classes of abuse work, including account creation, social networking link generation and search engine optimization support, and characterize how pricing and demand have evolved in supporting this activity. We show that scammers heavily employ outsourced labor, with abuse jobs constituting approximately 30% of the job solicitations on the site. Further, we demonstrate that workers quickly adapt their skill sets in responses to changes in demand for various abuse tasks. Lastly, the engagement portion of our study shows that workers actually deliver the promised goods, though the quality of the items is often variable. The second part of the dissertation focuses exclusively on the role of humans in circumventing CAPTCHAs. Human CAPTCHA solving services represent a heavily commercialized, outsourced abuse task, and we perform an in-depth analysis of this industry. CAPTCHAs are an ubiquitous defense used to protect open Web resources from being exploited at scale. In response to the widespread deployment of CAPTCHAs, a robust solving ecosystem has emerged, selling real-time human labor to bypass these protections. We analyze the behavior and dynamics of CAPTCHA-solving service providers, their price performance, and the underlying labor markets driving this economy. Ultimately, our work shows that CAPTCHAs are effective at differentiating between humans and computers. However, due to the vast number of human workers willing to solve CAPTCHAs for low wages, CAPTCHAs cannot necessarily prevent widespread abuse; instead, they serve as a low- cost economic impediment to abusers The results from these two studies demonstrate the increasing role that outsourcing plays in abusing Web services at scale. Furthermore, they suggest that Web services not only need to consider automated threats, but also must contend with an agile human labor pool. Lastly, they suggest one way to evaluate deployed security mechanisms, by monitoring the price and demand fluctuations for various abusively obtained products

Genetic Algorithms (GAs) make use of an internal representation of a given system in order to perform optimization functions. The actual structural layout of this representation, called a genome, has a crucial impact on the outcome of the optimization process. The purpose of this paper is to study the effects of different internal representations in a G A , which generates neural networks. A second G A was used to optimize the genome structure. This structure produces an optimized system within a shorter time interval.

This paper studies several applications of genetic algorithms (GAs) within the neural networks field. T he system was used to generate neural network circuit architectures. This was accomplished by using the G A to determine the weights in a fully interconnected network. The importance of the internal genetic representation was shown by testing different approaches. The effects in speed of optimization of varying the constraints imposed upon the desired network were also studied. It was observed that relatively loose constraints provided results comparable to a fully constrained system. The type of neural network circuits generated were recurrent competitive fields as described by Grossberg (1982).

For an instructor who is teaching a massive open online course (MOOC), what is the best way to understand their class? What is the best way to view how the students are interacting with the content while the course is running? To help prepare for the next iteration, how should the course's data be best analyzed after the fact? How do these instructional monitoring needs differ between online courses with tens of thousands of students and courses with only tens? This paper reports the results of a survey of 92 MOOC instructors who answered questions about which information they find useful in their course, with the end goal of creating an information display for MOOC instructors. The main findings are: (i) quantitative data sources such as grades, although useful, are not sufficient; understanding the activity in discussion forums and student surveys was rated useful for all use cases by a large majority of respondents, (ii) chat logs were not seen as useful, (iii) for the most part, the same sources of information were seen as useful as found in surveys of smaller online courses, (iv) mockups of existing and novel visualization techniques were responded to positively for use both while the course is running and for planning a revision of the course, and (v) a wide range of views was expressed about other details.

COVID-19 and the 2021 U.S. Capitol attacks have highlighted the potential dangers of pseudoscientific and conspiratorial belief adoption. Approaches to combating misinformed beliefs have tried to "pre-bunk" or "inoculate" people against misinformation adoption and have yielded only modest results. These approaches presume that some citizens may be more gullible than others and thus susceptible to multiple misinformed beliefs. We provide evidence of an alternative account: it's simply too hard for all people to be accurate in all domains of belief, but most individuals are trying. We collected data on a constellation of human beliefs across domains from more than 1,700 people on Amazon Mechanical Turk, and find misinformed beliefs to be broadly, but thinly, spread among the population. Further, we do not find that individuals who adopt one misinformed belief are more likely to engage in pseudoscientific or conspiratorial thinking across the board, in opposition to "slippery slope” notions of misinformation adoption.

A common task in qualitative data analysis is to characterize the usage of a linguistic entity by issuing queries over syntactic relations between words. Previous interfaces for searching over syntactic structures require programming-style queries. User interface research suggests that it is easier to recognize a pattern than to compose it from scratch; therefore, interfaces for non-experts should show previews of syntactic relations. What these previews should look like is an open question that we explored with a 400-participant Mechanical Turk experiment. We found that syntactic relations are recognized with 34% higher accuracy when contextual examples are shown than a baseline of naming the relations alone. This suggests that user interfaces should display contextual examples of syntactic relations to help users choose between different relations. © 2014 Association for Computational Linguistics.