AI Washing: The Framing Effect of Labels on Algorithmic Advice Utilization

2023 | International Conference on Information Systems | Citations: 0

Authors: Leffrang, Dirk; Mueller, Oliver

Abstract: Many researchers and practitioners see artificial intelligence as a game change ... Expand

Abstract: Many researchers and practitioners see artificial intelligence as a game changer compared to classical statistical models. However, some software providers engage in "AI washing", relabeling solutions that use simple statistical models as AI systems. By contrast, research on algorithm aversion unsystematically varied the labels for advisors and treated labels such as "artificial intelligence" and "statistical model" synonymously. This study investigates the effect of individual labels on users' actual advice utilization behavior. Through two incentivized online within-subjects experiments on regression tasks, we find that labeling human advisors with labels that suggest higher expertise leads to an increase in advice-taking, even though the content of the advice remains the same. In contrast, our results do not suggest such an expert effect for advice-taking from algorithms, despite differences in self-reported perception. These findings challenge the effectiveness of framing intelligent systems as AI-based systems and have important implications for both research and practice. Collapse

Topics: artificial intelligence decision making individual characteristic user experience user behavior

Methods: computational algorithm experiment survey longitudinal research within subject experiment

Exploiting Linked Open Data for a Collaborative Filtering Recommendation System

2023 | Americas Conference on Information Systems | Citations: 0

Authors: de Souza, Paulo; Durao, Frederico; Almeida, João Paulo Dias de; Mascarenhas, Ramon; Oliveira, Mayki Dos Santos

Abstract: Recommendation Systems (RS) tackle Information Overload on the web, and Collabo ... Expand

Abstract: Recommendation Systems (RS) tackle Information Overload on the web, and Collaborative Filtering (CF) is a widely used technique to estimate content relevance for users based on similar profiles. Traditional CF, however, often overlooks hidden semantic information and item relationships. This work proposes an RS that integrates Linked Open Data with CF to improve recommendation precision by leveraging semantic information about resources and their relationships. The approach calculates semantic similarity among resources using the Linked Open Data graph. By comparing error rates (MAE and RMSE) in varying scenarios, the proposed approach shows at least a 1.5% improvement in precision rates compared to traditional CF. This study contributes insights for researchers seeking to exploit Linked Open Data in recommendation contexts. Collapse

Topics: collaborative filtering open data recommender system database system knowledge representation

Methods: computational algorithm k-nearest neighbor DBpedia experiment literature study

Utilizing Adaptive AI-based Information Systems to Analyze the Effectiveness of Policy Measures in the Fight of COVID-19

2020 | International Conference on Information Systems | Citations: 0

Authors: Baier, Lucas; Schoeffer, Jakob; Kã, Niklas

Abstract: A novel coronavirus has been spreading around the world since early 2020. Due to ... Expand

Abstract: A novel coronavirus has been spreading around the world since early 2020. Due to its high infectiousness and lethality, countries around the world have adopted drastic policy measures to contain the pandemic—with far-reaching implications for economies and societies. We analyze the effect of two nonpharmaceutical interventions (NPI), namely closure of schools as well as a lockdown, on the development of daily case numbers of COVID-19 patients in several countries and US states. By applying Adaptive AIbased Information Systems (AAIS) based on concept drift detection, we conclude that it takes more than two weeks on average until NPIs show a significant effect on the number of new cases. This information is crucial for policymakers as they are currently confronted with difficult decisions to trade off the reopening of the economy with a strict containment of the virus. Collapse

Topics: artificial intelligence pandemic systems development accounting digital dashboard

Methods: computational algorithm time series analysis longitudinal research conceptual modelling regression analysis method

Modern Advanced Analytics Platforms and Predictive Models for Stock Price Forecasting: IBM Watson Analytics Case

2017 | HICSS | Citations: 0

Authors: Faizullov, Ilias; Yablonsky, Sergey

Abstract: The primary purpose of this paper was to provide an in-depth analysis of the abi ... Expand

Abstract: The primary purpose of this paper was to provide an in-depth analysis of the ability of modern analytical platforms (using IBM Watson Analytics as an example) to generate predictive models for stock prices forecasting in comparison with traditional analytical econometric platforms and models. Series of stock predictive models based on the suggestions of IBM Watson Analytics have demonstrated results, which are superior to all other models. In terms of forecasting accuracy, they beat all models except for the Random Walk. The simulation has demonstrated high returns for most of the suggested models. Collapse

Topics: analytical information system spreadsheet open source website

Methods: simulation regression analysis method ordinary least square experiment econometric model

Theories: transaction cost economics capital asset pricing model

The effect of data preprocessing on a retail price optimization system

2016 | Decision Support Systems | Citations: 1

Authors: Kunz, Timo P.; Crone, Sven F.; Meissner, Joern

Abstract: Revenue management (RM) is making a significant impact on pricing research and p ... Expand

Abstract: Revenue management (RM) is making a significant impact on pricing research and practice, from aviation and hospitality industries to retailing. However, empirical data conditions in retail are distinct to other industries, in particular in the large number of products within and across categories. To set profitable static prices with established RM models, the data is often simplified by data pruning (the exclusion of subsets of data that are deemed irrelevant or unsuitable) and data aggregation (the combination of disparate data points). However, the impact of such data preprocessing, despite being ubiquitous in retailing, is insufficiently considered in current RM research. This could induce potential sources of bias for the demand model estimates, as well as subsequent effects on the price optimization system, the optimized price set, and the profit maxima, which have not yet been investigated. This paper empirically studies the impact of two commonly used data preprocessing techniques in retail RM, data pruning and data aggregation, using simulated and empirical retail scanner data. We numerically assess potential biases introduced by data preprocessing using a systems perspective in estimating a two-stage demand model, the resulting price elasticities, optimized price sets, and the ensuing profit that it yields. Results show that both data aggregation and data pruning bias demand model estimates, albeit with different effect, but both produce less profitable price sets than unbiased reference solutions. The results demonstrate the practical importance of data preprocessing as a cause for estimation bias and suboptimal pricing in retail price optimization systems. Collapse

Topics: price management decision support system logistics management database system dynamic pricing

Methods: simulation optimization model data modeling descriptive statistic non linear optimization

The Ambiguous Identifier Clustering Technique

2016 | Electronic Markets | Citations: 3

Authors: Scholz, Michael; Franz, Markus; Hinz, Oliver

Abstract: Investigations of online transaction data often face the problem that entries fo ... Expand

Abstract: Investigations of online transaction data often face the problem that entries for identical products cannot be identified as such. There is, for example, typically no unique product identifier in online auctions; retailers make their offers at price comparison sites hardly comparable and online stores often use different identifiers for virtually equal products. Existing studies typically use data sets that are restricted to one or only a few products in order to avoid product heterogeneity if a unique product identifier is not available. We propose the Ambiguous Identifier Clustering Technique (AICT) that identifies online transaction data that refer to virtually the same product. Based on a data set of eBay auctions, we demonstrate that AICT clusters online transactions for identical products with high accuracy. We further show how researchers benefit from AICT and the reduced product heterogeneity when analyzing data with econometric models. Collapse

Topics: database system online auction marketing management application programming interface

Methods: cluster analysis agglomerative hierarchical clustering econometric modeling DBSCAN regression analysis method

Enabling Sustainable Smart Homes: An Intelligent Agent Approach

2014 | International Conference on Information Systems | Citations: 7

Authors: Valogianni, Konstantina; Ketter, Wolfgang; Collins, John

Abstract: Smart homes can offer significant benefits to residents. Specifically, their com ... Expand

Abstract: Smart homes can offer significant benefits to residents. Specifically, their combination with electric vehicles (EVs) can support environmental sustainability, as they can use part of their battery to cover household consumption needs. We present a Mobility Integrated Energy Management IS artifact that supports a smart home owner’s decisions with regard to using household appliances and charging electric vehicles. The artifact offers personalized energy consumption recommendations based on individual characteristics using information available. We observe that by adopting it, owners reshape their energy consumption curve and can save on their electricity bill. At the same time they create benefits for the electricity grid by reducing peak demand and increasing sustainability. We conclude by offering energy policy recommendations with regard to EV and smart home appliances adoption rates. Collapse

Topics: smart home accounting price management green IS global navigation satellite system

Methods: simulation design artifact autoregressive moving average model sensitivity analysis longitudinal research

Theories: theory of affordance