People often rely on heuristic reasoning when receiving algorithm advice, and this reliance leads to biased decisions that undermine the effectiveness of human-AI collaboration. Such bias persists even when individuals are given more time to deliberate or provided more information about AI, as they may lack the awareness or ability to engage in systematic reasoning. In this paper, we explore how guided reflection may enhance decision-making performance in human-AI collaboration by prompting a systematic reasoning process. We conducted an experiment with 178 participants, comparing decision-making behavior across three conditions: AI, explainable AI (XAI), and XAI with reflection. The results demonstrate that reflection significantly reduced over-reliance on AI and improved decision accuracy. Individuals with a high need for cognition and a high perceived understanding of AI benefited more from reflection. Furthermore, our study uncovers distinct patterns of cognitive processing and belief adjustment across different experimental conditions. Our findings provide a practical strategy for fostering cognitive engagement and contribute to a deeper understanding of human cognitive processes in AI-assisted decision-making.

Data quality is important to analytics; data preparation usually involves data cleaning and is often the most time-consuming part of analytics projects. When the topic is left to the discretion of individual courses in an analytics program, students often end up with light exposure to the topic. Instead, a course on data quality in analytics has been designed and implemented. Organized in eight modules, the first part of the course covers data preparation and preprocessing. This prepares students with the ability to tackle real datasets in other analytics courses. The second part covers analytics for data quality where algorithms for detecting and resolving data quality issues are covered. The third part addresses large scale and engineering issues of analytics practice where data collection needs to be managed and data quality tasks must be part of the pipeline.

It is a challenging task to predict financial markets. The complexity of this task is mainly due to the interaction between financial markets and market participants, who are not able to keep rational all the time, and often affected by emotions such as fear and ecstasy. Based on the state-of-the-art approach particularly for financial market predictions, a hybrid convolutional LSTM Based variational sequence-to-sequence model with attention (CLVSA), we propose a novel deep learning approach, named dual-CLVSA, to predict financial market movement with both trading data and the corresponding social sentiment measurements, each through a separate sequence-to-sequence channel. We evaluate the performance of our approach with backtesting on historical trading data of SPDR SP 500 Trust ETF over eight years. The experiment results show that dual-CLVSA can effectively fuse the two types of data, and verify that sentiment measurements are not only informative for financial market predictions, but they also contain extra profitable features to boost the performance of our predicting system.

Financial markets are a complex dynamical system. The complexity comes from the interaction between a market and its participants, in other words, the integrated outcome of activities of the entire participants determines the markets trend, while the markets trend affects activities of participants. These interwoven interactions make financial markets keep evolving. Inspired by stochastic recurrent models that successfully capture variability observed in natural sequential data such as speech and video, we propose CLVSA, a hybrid model that consists of stochastic recurrent networks, the sequence-to-sequence architecture, the self- and inter-attention mechanism, and convolutional LSTM units to capture variationally underlying features in raw financial trading data. Our model outperforms basic models, such as convolutional neural network, vanilla LSTM network, and sequence-to-sequence model with attention, based on backtesting results of six futures from January 2010 to December 2017. Our experimental results show that, by introducing an approximate posterior, CLVSA takes advantage of an extra regularizer based on the Kullback-Leibler divergence to prevent itself from overfitting traps.

We present a comprehensive classification of data misinterpretation problems and develop an approach to automatic detection and reconciliation of data interpretation conflicts in Web services composition. The approach uses a lightweight ontology augmented with modifiers, contexts, and atomic conversions between the contexts. The WSDL descriptions of Web services are annotated to establish correspondences to the ontology. Given the naive Business Process Execution Language (BPEL) specification of the desired Web services composition with data interpretation conflicts, the approach can automatically detect the conflicts and produce the corresponding mediated BPEL. Finally, we develop a prototype to validate and evaluate the approach.

Web services have become the technology of choice for service-oriented computing implementation, where Web services can be composed in response to some users' needs. It is critical to verify the compatibility of component Web services to ensure the correctness of the whole composition in which these components participate. Traditionally, two conditions need to be satisfied during the verification of compatibility: reachable termination and proper termination. Unfortunately, it is complex and time consuming to verify those two conditions. To reduce the complexity of this verification, we model Web services using colored Petri nets (PNs) so that a specific property of their structures is looked into, namely, well structuredness. We prove that only reachable termination needs to be satisfied when verifying behavioral compatibility among well-structured Web services. When a composition is declared as valid and in the case where one of its component Web services fails at run time, an alternative one with similar behavior needs to come into play as a substitute. Thus, it is important to develop effective approaches that permit one to analyze the similarity of Web services. Although many existing approaches utilize PNs to analyze behavioral compatibility, few of them explore further appropriate definitions of behavioral similarity and provide a user-friendly tool with automatic verification. In this paper, we introduce a formal definition of context-independent similarity and show that a Web service can be substituted by an alternative peer of similar behavior without intervening other Web services in the composition. Therefore, the cost of verifying service substitutability is largely reduced. We also provide an algorithm for the verification and implement it in a tool. Using the tool, the verification of behavioral similarity of Web services can be performed in an automatic way.

The potential benefits of Web services composition heavily rely on semantic interoperability, i.e., the ability to exchange data meaningfully amongst Web services. Context heterogeneity, which refers to different implicit assumptions about interpreting the exchanged data, hampers the automatic composition of Web services. However, existing initiatives of semantic Web services (SWSs) often ignore context heterogeneity. In this paper, we introduce an approach to address this issue. The contexts of the involved Web services are defined in a lightweight ontology and their WSDL descriptions are annotated by an extension of a W3C standard, i.e., semantic annotation for WSDL and XML schema (SAWSDL). The composition of Web services is described using BPEL specification. Given a BPEL file that ignores context heterogeneity, the approach automatically detects all context differences among the involved services, and reconciles them by producing a mediated BPEL file that incorporates necessary conversions using Xpath functions and/or Web services.

The change in meaning of data over time poses significant challenges for the use of that data. These challenges exist in the use of an individual data source and are further compounded with the integration of multiple sources. In this paper, we identify three types of temporal semantic heterogeneities. We propose a solution based on extensions to the context interchange framework, which has mechanisms for capturing semantics using ontology and temporal context. It also provides a mediation service that automatically resolves semantic conflicts. We show the feasibility of this approach with a prototype that implements a subset of the proposed extensions.