@inproceedings{37312, abstract = {{Optimal decision making requires appropriate evaluation of advice. Recent literature reports that algorithm aversion reduces the effectiveness of predictive algorithms. However, it remains unclear how people recover from bad advice given by an otherwise good advisor. Previous work has focused on algorithm aversion at a single time point. We extend this work by examining successive decisions in a time series forecasting task using an online between-subjects experiment (N = 87). Our empirical results do not confirm algorithm aversion immediately after bad advice. The estimated effect suggests an increasing algorithm appreciation over time. Our work extends the current knowledge on algorithm aversion with insights into how weight on advice is adjusted over consecutive tasks. Since most forecasting tasks are not one-off decisions, this also has implications for practitioners.}}, author = {{Leffrang, Dirk and Bösch, Kevin and Müller, Oliver}}, booktitle = {{Hawaii International Conference on System Sciences}}, keywords = {{Algorithm aversion, Time series, Decision making, Advice taking, Forecasting}}, title = {{{Do People Recover from Algorithm Aversion? An Experimental Study of Algorithm Aversion over Time}}}, year = {{2023}}, } @inproceedings{17810, abstract = {{In all fields, the significance of a reliable and accurate predictive model is almost unquantifiable. With deep domain knowledge, models derived from first principles typically outperforms other models in terms of reliability and accuracy. When it may become a cumbersome or an unachievable task to build or validate such models of complex (non-linear) systems, machine learning techniques are employed to build predictive models. However, the accuracy of such techniques is not only dependent on the hyper-parameters of the chosen algorithm, but also on the amount and quality of data. This paper investigates the application of classical time series forecasting approaches for the reliable prognostics of technical systems, where black box machine learning techniques might not successfully be employed given insufficient amount of data and where first principles models are infeasible due to lack of domain specific data. Forecasting by analogy, forecasting by analytical function fitting, an exponential smoothing forecasting method and the long short-term memory (LSTM) are evaluated and compared against the ground truth data. As a case study, the methods are applied to predict future crack lengths of riveted aluminium plates under cyclic loading. The performance of the predictive models is evaluated based on error metrics leading to a proposal of when to apply which forecasting approach.}}, author = {{Aimiyekagbon, Osarenren Kennedy and Bender, Amelie and Sextro, Walter}}, booktitle = {{PHM Society European Conference}}, keywords = {{PHM 2019, crack propagation, forecasting, unevenly spaced time series, step ahead prediction, short time series}}, number = {{1}}, title = {{{Evaluation of time series forecasting approaches for the reliable crack length prediction of riveted aluminium plates given insufficient data}}}, volume = {{5}}, year = {{2020}}, } @techreport{20869, abstract = {{This study provides evidence of significant biases in multi-year management forecasts by analyzing a proprietary dataset on venture-backed start-ups in Germany. We find that revenues and expenses are highly overestimated in each of the investigated one- to five-year-ahead planning periods. Furthermore, entrepreneurs underestimate one-year-ahead profit forecasts but clearly overestimate their profit forecasts for all longer-term forecast horizons. Additional analyses reveal that teams with prior management experience issue even more overestimated forecasts and misrepresent their forward-looking information. In contrast, greater asset verifiability and corporate lead investors are associated with lower levels of forecast errors. All key results hold if bias is either measured by traditionally comparing forecasts to ex-post realizations or by using a cross-sectional projection approach based on historical accounting data developed by prior research.}}, author = {{Sievers, Sönke and Mokwa, Christopher Frederik}}, keywords = {{Management forecasts, Forecasting biases, Venture-backed start-ups, Projection methods}}, pages = {{42}}, title = {{{Biases in Management Forecasts of Venture-Backed Start-Ups: Evidence from Internal Due Diligence Documents of VC Investors}}}, doi = {{10.2139/ssrn.1714399}}, year = {{2012}}, } @article{5198, abstract = {{This study provides evidence of significant biases in multi-year management forecasts by analyzing a proprietary dataset on venture-backed start-ups in Germany. We find that revenues and expenses are highly overestimated in each of the investigated one- to five-year-ahead planning periods. Furthermore, entrepreneurs underestimate one-year-ahead profit forecasts but clearly overestimate their profit forecasts for all longer-term forecast horizons. Additional analyses reveal that teams with prior management experience issue even more overestimated forecasts and misrepresent their forward-looking information. In contrast, greater asset verifiability and corporate lead investors are associated with lower levels of forecast errors. All key results hold if bias is either measured by traditionally comparing forecasts to ex-post realizations or by using a cross-sectional projection approach based on historical accounting data developed by prior research. }}, author = {{Mokwa, Christopher and Sievers, Sönke}}, journal = {{SSRN Electronic Journal}}, keywords = {{Management forecasts, Forecasting biases, Venture-backed start-ups, Projection methods}}, title = {{{Biases in Management Forecasts of Venture-Backed Start-Ups: Evidence from Internal Due Diligence Documents of VC Investors}}}, doi = {{10.2139/ssrn.1714399}}, year = {{2012}}, }