MWMarkus Weinmann

Markus Weinmann

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Markus Weinmann has a total of 635 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Can Human Verification Restore Felt Responsibility for AI-Assisted Knowledge Work?

    by Stefan Rose, Markus Weinmann

    Generative AI (genAI) co-produces the work of hundreds of millions of people. Humanoversight, the default safeguard in AI governance, assumes that people still feel responsiblefor the outputs they approve. Yet genAI assistance can lower psychological ownership ofsuch outputs, and whether the resulting loss of felt responsibility can be repaired remainsunclear. Drawing on the triangle model of responsibility and psychological ownership theory,we propose that psychological ownership is how people experience the causal link betweenthemselves and an outcome, that genAI assistance weakens this link, and that a mandatoryverification step after generation can rebuild it. We derive five hypotheses on the erosion andrestoration of psychological ownership and felt responsibility and on revision after criticism,and we outline a 3 × 2 experiment to test them.

  • Open Access

    Causal machine learning reveals heterogeneous effects of loss versus gain framing on retirement savings

    by Kevin Bösch, Oliver Müller, Markus Weinmann, Stefan Feuerriegel

    As Western countries face rapidly aging populations, insufficient retirement savings threaten individual financial security and the sustainability of social welfare systems. Behavioral interventions such as loss–gain message framing are often proposed as scalable solutions, yet evidence on their effectiveness is mixed, largely because little is known about when and for whom they work. We analyze heterogeneity in the effects of loss–gain framing on real-life, high-stakes retirement saving decisions. In a large-scale randomized field experiment ( $$N = 22{,}919$$ ) with a European retirement savings company, individuals were randomized to loss- or gain-framed reminders designed to increase participation (extensive margin) and savings amount (intensive margin). Using causal machine learning, we estimate heterogeneous treatment effects across socioeconomic characteristics and validate them in a second, independent experiment ( $$N = 18{,}288$$ ). In this validation experiment, for the 25% most responsive to loss framing, conversion is estimated to rise by 9.4 percentage points relative to gain framing, whereas for the 25% most responsive to gain framing it is expected to fall by 8.4 points. For savings amount, we estimate that loss framing raises contributions for 64% of contributors (by €737 on average) but backfires for the rest (by €615 on average). Overall, a personalized targeting policy based on causal machine learning is expected to increase participation by 51% (from 4.53% to 6.86%) and average savings by 18% (from €2,062 to €2,442 per converted customer) relative to uniform loss framing.

  • Open Access

    Automated reproducibility assessments in the social and behavioral sciences using large language models

    by Stefan Feuerriegel, Tobias Holtdirk, Pietro Marcolongo, Anna Steinberg Schulten, Felix Henninger, Stefan Rose, Sarah Ball, Bolei Ma, Frauke Kreuter, Markus Weinmann

    Reproducibility in the social and behavioral sciences is typically evaluated by independent researchers who reanalyze the original data to assess whether the published findings can be recovered. However, such approaches are resource-intensive and difficult to scale. Here, we show that large language models (LLMs) can automate reproducibility assessments. Using N=180 published studies with predefined claims from the behavioral and social sciences, we compare LLM-generated analyses with the original findings. For 11 studies, the LLM pipeline could not produce a viable effect size estimate. For the remaining studies, the LLM reached the same qualitative conclusion as the original study in 80% of cases, and recovered the original effect sizes (using a +/-0.05 tolerance in Cohen's d) in 24% of studies. In a subset with human reanalyses, the LLM reached the same qualitative conclusion as the original study in 95% of studies, similar to human reanalysts (83%), and the LLM recovered the original effect sizes using a $\pm$0.05 tolerance in 40% of studies, again broadly similar to human reanalysts (28%). Given the current capabilities and limitations of LLMs, the findings show that LLMs can support systematic audits of empirical results rather than substitute expert judgment. As such, LLMs can serve as a scalable screening tool to improve the rigor and reproducibility in empirical research.

  • Open Access

    Machine learning reveals heterogeneity in the effectiveness of democracy-strengthening interventions

    by Stefan Feuerriegel, Til Gerlach, Abdurahman Maarouf, Markus Weinmann

    Anti-democratic attitudes (ADA) threaten the stability of democratic systems. Several behavioral interventions have been proposed to strengthen democratic attitudes, yet the heterogeneity in their effectiveness remains unclear. In particular, it is unknown for whom these interventions are most effective, raising concerns that uniform approaches may benefit some groups while leaving others behind. Here, we use causal machine learning to analyze individual-level differences in the effectiveness of 25 behavioral interventions aimed to reduce ADA using data from a large-scale randomized controlled trial (N = 32,059). First, we uncover substantial heterogeneity of the interventions. For example, even the least-effective intervention still reduces ADA for 28% of individuals. Second, we identify key drivers across demographic and ideological characteristics, finding that political ideology is a strong, consistent predictor, even across party lines. For example, interventions based on factual correction are more effective for conservatives, while interventions appealing to institutional trust are more effective for liberals. Our findings highlight the limitations of one-size-fits-all strategies in reducing ADA and emphasize the importance of accounting for heterogeneity to maximize the effectiveness of democracy-strengthening interventions. Overall, these findings raise important equity concerns, as uniform intervention strategies may disproportionately benefit some groups but may systematically fail to reach vulnerable subgroups.

  • Open Access

    Investigating the analytical robustness of the social and behavioural sciences

    by Balazs Aczel, Barnabas Szaszi, Harry T Clelland, Marton Kovacs, Felix Holzmeister, Don van Ravenzwaaij, Hannah Schulz-Kümpel, Sabine Hoffmann, Gustav Nilsonne, Livia Kósa, et al.

    The same dataset can be analysed in different justifiable ways to answer the same research question, potentially challenging the robustness of empirical science1–3. In this crowd initiative, we investigated the degree to which research findings in the social and behavioural sciences are contingent on analysts’ choices. We examined a stratified random sample of 100 studies published between 2009 and 2018, where for one claim per study, at least five re-analysts independently re-analysed the original data. The statistical appropriateness of the re-analyses was assessed in peer evaluations, and the robustness indicators were inspected along a range of research characteristics and study designs. We found that 34% of the independent re-analyses yielded the same result (within a tolerance region of +/- 0.05 Cohen’s d) as the original report; with a four times broader tolerance region, this indicator rose to 57%. Regarding the conclusions drawn, 74% of analyses were reported to arrive at the same conclusion as in the original investigation; 24% to no effects/inconclusive result, and 2% to the opposite effect as in the original investigation. This exploratory study suggests that the common single-path analyses in social and behavioural research should not simply be assumed to be robust to alternative analyses4. Therefore, we recommend the development and use of practices to explore and communicate this neglected source of uncertainty.

  • Open Access

    Beyond means: A dynamic framework for predicting customer satisfaction

    by Christof Naumzik, Abdurahman Maarouf, Stefan Feuerriegel, Markus Weinmann


  • Open Access

    Analyzing heterogeneity in the effects of behavioral climate interventions on individual beliefs across sociodemographics, attitudes, and norms

    by Stefan Feuerriegel, Abdurahman Maarouf, Sercan Demir, Markus Weinmann

    Addressing climate change requires global shifts in individual beliefs at scale. While several behavioral interventions have been developed to strengthen climate change beliefs (e.g., by framing climate change as a psychologically proximal risk), the effectiveness of such interventions differs across individuals. Yet, the extent and drivers of such heterogeneity in intervention effectiveness remain largely unclear. Here, we aim to understand individual-level differences in the effectiveness of 11 expert-crowdsourced interventions on climate change beliefs across two randomized controlled trials (RCTs) (RCT 1: N = 59,508 individuals from 63 countries; RCT 2: N = 1,610). First, our analysis reveals large heterogeneity in the effectiveness: even the most effective intervention in terms of average treatment effect fails to motivate belief change for 20% of individuals, while interventions regarded as “ineffective” still positively influence 47%. Second, we identify key drivers across sociodemographics, attitudes, and norms, thereby showing that interventions effective in the Global North often backfire in the Global South, where different socioeconomic contexts shape responses. Third, we performed an online field study (N = 772,777) to externally validate the findings, showing that personalizing interventions increases effectiveness by more than 30%. Our findings emphasize the importance of personalization in behavioral climate interventions to maximize the effectiveness of behavioral strategies and foster widespread support of climate action.


  • The Digital Experiment Reporting Protocol (DERP)

    by University of Cologne, Markus Weinmann, Christoph Schneider, Washington State University, Joseph Valacich, University of Arizona

    Transparent reporting is a cornerstone of credible and reproducible research, yet digital experiments—randomized studies conducted in online or computerized environments—often lack clear reporting standards. To address this gap, we adapt and extend the CONSORT Statement, a widely used checklist for clinical trials, to develop a reporting checklist tailored for digital experiments. While the checklist is broadly applicable across disciplines, we demonstrate its relevance through an empirical application to digital experiments published in two leading Information Systems (IS) journals: Information Systems Research and MIS Quarterly. Our analysis reveals substantial shortcomings in reporting quality, with only 43% of the papers providing sufficient detail on core items. These findings highlight the urgent need for standardized reporting in digital experimentation. The proposed checklist offers a practical tool for authors, reviewers, and editors to enhance the transparency, rigor, and replicability of digital experiments.

  • Open Access

    Message framing to promote solar panels

    by Dominik Bär, Stefan Feuerriegel, Ting Li, Markus Weinmann

    Green technologies, such as solar panels, foster the use of clean energy, yet often involve large-scale investments. Hence, adoption by retail consumers has been a key barrier. Here, we show that message framing can significantly increase customers’ serious commitment to adopting solar panels by providing empirical evidence in the field from a large-scale randomized controlled trial with a nationwide online retailer in the Netherlands (N = 26,873 participants). We design four messages aimed at promoting the purchase behavior of solar panel installations. Our messages present outcomes for oneself or for the environment and highlight cost savings versus earnings (for oneself) or reducing emissions versus generating green electricity (for the environment). Across all messages, we observe a higher rate of customers committing to solar panels compared to the baseline. However, the framing in terms of financial savings for oneself was by far the most effective, resulting in a 40% higher level of commitment than the baseline and 30% higher than the average of the other three messages, which were not significantly different in effect from each other. Our results show that message framing is cost-efficient and scalable among retail consumers to promote large-scale investments in green technologies and thus clean energy.


  • The Attraction Effect in Crowdfunding

    by Markus Weinmann, Abhay Nath Mishra, Lena Franziska Kaiser, Jan Vom Brocke

    Crowdfunding has reached enormous popularity among entrepreneurs, yet many struggle to fund their projects. For example, only 39.8% of all Kickstarter projects get funded. Why? Entrepreneurs need to set a predefined funding goal, which needs to be reached to get the project funded. If not reached, they do not get the money. To reach the funding goal, entrepreneurs can design reward menus from which backers can choose low-priced to high-priced rewards to support a given project. We show that simply inserting a so-called decoy option—an option in the first place that does not make sense but makes higher-priced options more attractive—can lead to significantly more backers choosing a high-priced option (up to 28% more high-priced choices). This makes it more likely for entrepreneurs to reach their predefined funding goals and thus start their designated projects.