Through four essays, this dissertation contributes to the body of marketing literature by advancing understanding of consumer interactions with digital decision aids. Different aspects of the content contained within digital decision aids are explored in several contexts. First, the drivers of consumer interactivity in an online review community are examined and it is found that violations of community norms are an important factor in stimulating consumer action. Second, a tool is developed to facilitate the normalization of online review content across languages. Next, elements of language and national culture are investigated to determine their influence on consumer reviews in an international context. It is found that cultural biases play an important role in the relative verbosity, valence, and helpfulness of online reviews across countries. Lastly, the role of images in digital decision aids is considered and it is found that image type and perspective can influence consumer product evaluation. In sum, the influence that content and context have on consumer interactions with digital decision aids is clearly demonstrated through a diverse yet intertwined set of studies. / Business Administration/Marketing

In this dissertation, I investigate how online social interactions and user-generated content affect sellers and consumers in online platforms. I conduct three empirical studies to understand the effect of user-generated content in three different types of online platforms: (1) an e-commerce marketplace, (2) an online reviews platform, and (3) an online health community. In study one, I examine how social features (e.g., following others, sharing others’ products) within an electronic commerce marketplace affect status and sales for sellers. This essay contributes to the literature on electronic commerce by deepening the understanding of online social processes among sellers. In study two, I explore how humorous appropriation of an online review platform affects purchase intention and consumer engagement. Utilizing both controlled experiments and analysis of real-world reviews, I demonstrate that humorous appropriation attenuates the effect of review valence on purchase intentions and increases consumer engagement. In study three, I investigate how community ratings are related to patient treatment evaluations and compliance in an online health community. I find that community ratings are positively associated with treatment evaluations and compliance. Moreover, I find that community size and ratings variance moderate the effect of community ratings on treatment evaluations and compliance. Taken together, these essays contribute to the literature on Information Systems by augmenting the understanding of the effects of different types of user-generated content on social (status, engagement, and evaluations) and economic outcomes (purchase intentions and sales). The studies also offer insights for strategic decisions regarding user-generated content in online platforms. / Business Administration/Management Information Systems

Systematic reviews (SR) and meta-analysis (MA) of randomised controlled trials (RCT) are the trustworthy sources of evidence. However, most systematic reviews focus on pair-wise comparisons. Network-meta-analysis (NMA) offers quantitative methods of integrating data from all the available comparisons of many different treatments for each outcome. In a systematic review of interventions, Summary of Findings (SoF) tables present the main findings of a review in a transparent and simple form. However, it is unknown how to present NMA findings in a tabular format. Moreover, systematic reviews and meta-analysis of interventions can summarize bodies of evidence from randomized and non-randomized studies (NRS). Integrating both sources of evidence in a single study can be challenging particularly in the context of assessing the certainty of the evidence, as well as presenting findings of both RCTs and NRS sources of evidence. In our study, we described how 276 NMA were conducted and how authors reported their main findings. We also conducted 32 interviews with users of NMAs and we designed two final NMA-SoF tables. Furthermore, we conducted two systematic reviews that included RCTs and NRS to address methodological challenges. Based on our results, we developed two NMA-SoF table formats to report the main findings of NMAs. The final format was appealing for users and allowed them to better understand NMA findings. Assessment of quality of individual NRS remains challenging and further research is needed to increase its appropriateness in systematic reviews of NRS. We determined that quality assessment of individual NRS was particularly challenging to implement due to the complexity of NRS evaluation tools. Our evaluation revealed that effect estimates of RCTs and NRS were better presented separately. / Thesis / Doctor of Philosophy (PhD) / Systematic reviews (SR) are a summary of studies that address a particular clinical question. Frequently, SRs are complemented with a statistical aggregation of results of individual studies to produce a single estimate. Summary of findings (SoF) tables are designed to present the most relevant information of systematic reviews and meta-analysis. However, it is unknown how to present network meta-analysis (NMA) findings in SoF tables. Another challenge relates to the integration of randomized controlled trials (RCTs) and non-randomized (NRS) studies. Methodological challenges in systematic reviews need to be addressed through careful research. In our study, we appraised how NMA were conducted, and how they presented their main findings. We designed two versions of SoF tables to present NMA findings. Moreover, we conducted two systematic reviews that included RCTs and NRS to address potential challenges in analyzing and presenting their findings.

This thesis explores the Rapid Recommendations process, a new responsive way of creating clinical practice guidelines. / Healthcare workers rely on clinical practice guidelines to inform their practice. However, most guidelines are not trustworthy when judged by accepted standards and they typically take several years to produce. Guideline trustworthiness is undermined by panel members who often have conflicts of interest, by including representation from only a subset of stakeholders, by failing to examine the entirety of the evidence systematically, and by rapid obsolescence. Further, they are often difficult for users to understand in limited time. Rather than updating guidance on a fixed schedule, the Rapid Recommendations approach involves continuous monitoring of the literature and produces guidelines in response to new potentially practice-changing evidence. A collaborative network of clinicians, methodologists, and patients respond by rapidly producing trustworthy evidence syntheses and guidance. We have identified efficiencies at every step of the guideline development process. The guideline panel does not include anyone with a financial conflict of interest and there are strict limits professional and intellectual conflicts. Systematic reviews are produced on the relative effects of each option, on prognosis, and on patient values and preferences with the explicit intent to inform the question at hand. The panel also considers practical issues. Rapid Recommendations are published in a concise multilayered user-friendly format headed by an interactive infographic that contains all of the necessary information for users need to make informed decisions at the point of care. The guideline is published simultaneously in print and electronically, including decision aids that can be used at the point of care and integrated into electronic medical records. In this thesis, you will find a selection of exemplary publications relevant to the Rapid Recommendations process. We show that a responsive approach to rapid and trustworthy guideline creation is possible. It represents a way forward from the current limitations that plague most current clinical practice guidelines. / Thesis / Candidate in Philosophy / Healthcare workers often decide what to do in practice based on the advice of experts through clinical practice guidelines. However, most clinical practice guidelines are not completely trustworthy. Guideline authors often have conflicts of interest, do not include patients or patient views, and are created so slowly that they rapidly fall into obsolescence. This thesis explores a new way of developing clinical practice guidelines that we call Rapid Recommendations. Instead of creating them on a fixed schedule (i.e., every few years), they are created in response to new studies that might change practice. The scope is limited, and timelines are shorter, meaning that the guidelines are published sooner. The guideline authors include all stakeholders, including patients. None of the authors have any financial interests in the topic, and other conflicts are minimized. The guidelines are published on an expedited basis and in an accessible online multilayered format with infographics. This thesis includes a selection of exemplary publications relevant to the Rapid Recommendations process.

The purpose of this dissertation intends to discover as well as categorize safety concern reports in online reviews by using key terms prevalent in sub-categories of safety concerns. This dissertation extends the literature of semi-automatic text classification methodology in monitoring and classifying product quality and service concerns. We develop various text classification methods for finding key concerns across a diverse set of product and service categories. Additionally, we generalize our results by testing the performance of our methodologies on online reviews collected from two different data sources (Amazon product reviews and Facebook hospital service reviews). Stakeholders such as product designers and safety regulators can use the semi-automatic classification procedure to subcategorize safety concerns by injury type and narrative type (Chapter 1). We enhance the text classification approach by proposing a Risk Assessment Model for quality management (QM) professionals, safety regulators, and product designers to allow them to estimate overall risk level of specific products by analyzing consumer-generated content in online reviews (Chapter 2). Monitoring and prioritizing the hazard risk levels of products will help the stakeholders to make appropriate actions on mitigating the risk of product safety. Lastly, the text classification approach discovers and ranks aspects of services that predict overall user satisfaction (Chapter 3). The key service terms are beneficial for healthcare providers to rapidly trace specific service concerns for improving the hospital services. / Doctor of Philosophy / This dissertation extends past studies by examining safety surveillance of online reviews. We examine online reviews reporting specific categories of safety concerns and contrast them with reviews not reporting these specific safety concerns. Business and regulators are benefited in detecting, categorizing, and prioritizing safety concerns across product categories. We use key terms prevalent in domain-related safety concerns for granular analysis of consumer reviews. Secondly, beyond utilizing the key terms to discover specific hazard incidents, safety regulators and manufacturers may use the extended risk assessment framework to estimate the risk severity, risk likelihood, and overall risk level of a specific product. The model could be useful for product safety practitioners in product risk identification and mitigation. Finally, this dissertation identifies the aspects of service quality concerns present in online hospital reviews. This study uses text analytics method by using key terms to detect these specific service concerns and hence determine primary rationales for patient feedback on hospital services. Managerially, this information helps to prioritize the areas in greatest need of improvement of hospital services. Additionally, generating key terms for a particular service attribute aids health care policy makers and providers in rapidly monitoring specific concerns and adjusting policies or resources to better serve patient

Due to the rapid development of Internet communication technologies (ICTs), an increasing number of social media platforms exist where consumers can exchange comments online about products and services that businesses offer. The existing literature has demonstrated that online user-generated content can significantly influence consumer behavior and increase sales. However, its impact on organizational operations has been primarily focused on marketing, with other areas understudied. Hence, there is a pressing need to design a research framework that explores the impact of online user-generated content on important organizational operations such as product innovation, customer relationship management, and operations management. Research efforts in this dissertation center on exploring the co-creation value of online consumer reviews, where consumers' demands influence firms' decision-making. The dissertation is composed of three studies. The first study finds empirical evidence that quality signals in online product reviews are predictors of the timing of firms' incremental innovation. Guided by the product differentiation theory, the second study examines how companies' innovation and marketing differentiation strategies influence app performance. The last study proposes a novel text analytics framework to discover different information types from user reviews. The research contributes theoretical and practical insights to consumer-centric innovation and social media analytics literature. / PHD

In recent years, the expansion of online media has presented firms with rich and voluminous new datasets with profound business applications. Among these, online reviews provide nuanced details on consumers' interactions with products. Analysis of these reviews has enormous potential, but the enormity of the data and the nature of unstructured text make mining these insights challenging and time-consuming. This paper presents three studies examining this problem and suggesting techniques for automated extraction of vital insights. The first study examines the problem of identifying mentions of safety hazards in online reviews. Discussions of hazards may have profound importance for firms and regulators as they seek to protect consumers. However, as most online reviews do not pertain to safety hazards, identifying this small portion of reviews is a challenging problem. Much of the literature in this domain focuses on selecting "smoke terms," or specific words and phrases closely associated with the mentions of safety hazards. We first examine and evaluate prior techniques to identify these reviews, which incorporate substantial human opinion in curating smoke terms and thus vary in their effectiveness. We propose a new automated method that utilizes a heuristic to curate smoke terms, and we find that this method is far more efficient than the human-driven techniques. Finally, we incorporate consumers' star ratings in our analysis, further improving prediction of safety hazard-related discussions. The second study examines the identification of consumer-sourced innovation ideas and opportunities from online reviews. We build upon a widely-accepted attribute mapping framework from the entrepreneurship literature for evaluating and comparing product attributes. We first adapt this framework for use in the analysis of online reviews. Then, we develop analytical techniques based on smoke terms for automated identification of innovation opportunities mentioned in online reviews. These techniques can be used to profile products as to attributes that affect or have the potential to affect their competitive standing. In collaboration with a large countertop appliances manufacturer, we assess and validate the usefulness of these suggestions, tying together the theoretical value of the attribute mapping framework and the practical value of identifying innovation-related discussions in online reviews. The third study addresses safety hazard monitoring for use cases in which a higher yield of safety hazards detected is desirable. We note a trade-off between the efficiency of hazard techniques described in the first study and the depth of such techniques, as a high proportion of identified records refer to true hazards, but several important hazards may be undetected. We suggest several techniques for handling this trade-off, including alternate objective functions for heuristics and fuzzy term matching, which improve the total yield. We examine the efficacy of each of these techniques and contrast their merits with past techniques. Finally, we test the capability of these methods to generalize to online reviews across different product categories. / Doctor of Philosophy / This dissertation presents three studies that utilize text analytic methods to analyze and derive insights from online reviews. The first study aims to detect distinctive words and phrases particularly prevalent in online reviews that describe safety hazards. This study proposes algorithmic and heuristic methods for identifying words and phrases that are especially common in these reviews, allowing for an automated process to prioritize these reviews for practitioners more efficiently. The second study extends these methods for use in detecting mentions of product innovation opportunities in online reviews. We show that these techniques can used to profile products based on attributes that differentiate them from competition or have the potential to do so in the future. Additionally, we validate that product managers find this attribute profiling useful to their innovation processes. Finally, the third study examines automated safety hazard monitoring for situations in which the yield or total number of safety hazards detected is an important consideration in addition to efficiency. We propose a variety of new techniques for handling these situations and contrast them with the techniques used in prior studies. Lastly, we test these methods across diverse product categories.

Yes / Online shopping is increasingly becoming people's first choice when shopping, as it is very convenient to choose products based on their reviews. Even for moderately popular products, there are thousands of reviews constantly being posted on e-commerce sites. Such a large volume of data constantly being generated can be considered as a big data challenge for both online businesses and consumers. That makes it difficult for buyers to go through all the reviews to make purchase decisions. In this research, we have developed models based on machine learning that can predict the helpfulness of the consumer reviews using several textual features such as polarity, subjectivity, entropy, and reading ease. The model will automatically assign helpfulness values to an initial review as soon as it is posted on the website so that the review gets a fair chance of being viewed by other buyers. The results of this study will help buyers to write better reviews and thereby assist other buyers in making their purchase decisions, as well as help businesses to improve their websites.

Yes / Product reviews are posted online by the hundreds and thousands for popular products. Handling such a large volume of continuously generated online content is a challenging task for buyers, sellers and researchers. The purpose of this study is to rank the overwhelming number of reviews using their predicted helpfulness scores. The helpfulness score is predicted using features extracted from review text, product description, and customer question-answer data of a product using the random-forest classifier and gradient boosting regressor. The system classifies reviews into low or high quality with the random-forest classifier. The helpfulness scores of the high-quality reviews are only predicted using the gradient boosting regressor. The helpfulness scores of the low-quality reviews are not calculated because they are never going to be in the top k reviews. They are just added at the end of the review list to the review-listing website. The proposed system provides fair review placement on review listing pages and makes all high-quality reviews visible to customers on the top. The experimental results on data from two popular Indian e-commerce websites validate our claim, as 3–4 newer high-quality reviews are placed in the top ten reviews along with 5–6 older reviews based on review helpfulness. Our findings indicate that inclusion of features from product description data and customer question-answer data improves the prediction accuracy of the helpfulness score. / Ministry of Electronics and Information Technology (MeitY), Government of India for financial support during research work through “Visvesvaraya PhD Scheme for Electronics and IT”.

Today many companies exist and market their products and services on social medias, and therefore may receive reviews and thoughts from their end-users directly in these social medias. Reading every text by hand can be time-consuming, so by analysing the sentiment for all texts give the companies an overview how positive or negative the users are on a specific subject. Sentiment analysis is a feature that Beanloop AB is interested in implementing in their future projects and this thesis research problem was to investigate how deep learning could be used for this task. It was done by conducting an experiment with deep learning and neural networks. Several convolutional neural network models were implemented with different settings to find a combination of settings that gave the highest accuracy on the given test dataset. There were two different kind of models, one kind classifying positive and negative, and the second classified the previous two categories but also neutral. The training dataset and the test dataset contained data from two recommendation sites, www.reco.se and se.trustpilot.com. The final result shows that when classifying three categories (positive, negative and neutral) the models had problems to reach an accuracy at 85%, were only one model reached 80% accuracy as best on the test dataset. However, when only classifying two categories (positive and negative) the models showed very good results and reached almost 95% accuracy for every model.