Third-party funding:
Individual social responsibility, bank risk, and the impact of regulation. Diligentia Foundation.
Co-author: Melina Ludolph.
Graduate Program "Navigating the Chaos of Innovation and Transformation", funded by the state Sachsen-Anhalt and the ESF+.
Jointly with Stefanie Börner, Michael Dick, Kai Heinrich, Elmar Lukas, Heike Ohlbrecht, Kersten Sven Roth and Marlin Ulmer.
Work in Progress:
Understanding citizens’ reactions to fiscal policy adjustments
Co-authors: Melina Ludolph and Lena Tonzer
Risk-taking on behalf of others: The role of professional identity and individual social responsibility
Co-authors: Melina Ludolph and Jatin Taneja
Talent Allocation to Startups versus Incumbents: A Cross-Country Experiment for Germany and the United States
Co-authors: Ufuk Akcigit, Javier Miranda, Steffen Müller, Patrick Nüß and Simon Wiederhold
Not All Co-Creation Is Equal: How Human and AI Involvement Shape Audience Perceptions of Artists
Co-authors: Kai Heinrich and Stefanie Kühn
As AI becomes increasingly involved in creative processes, audiences are confronted with provenance information about who, or what, contributed to a creative work. This raises a central question for human-AI co-creation: do audiences discount distributed authorship as such, or do they discount specifically technological co-creation? Because human co-creators and AI tools occupy the same external side of attribution theory's locus-of-causality dimension, the theory as commonly applied predicts that both should dilute audiences' artist-directed attributions. Drawing on Kelley's discounting principle and correspondent inference theory, we develop a source-dependent discounting account that predicts an asymmetry ex ante: human collaboration carries little information and triggers weak discounting of the artist's internal qualities, whereas AI involvement shifts causal credit outside the boundary of human expressive agency and triggers strong discounting. We test this account in a pre-registered online experiment with 1,381 participants, combining a solo-artist baseline with a 2 (co-creator source: human vs. AI) x 3 (stage: ideation, realization, both) factorial structure. Equivalence tests show that human co-creation leaves artist perception statistically equivalent to solo creation, whereas stage-matched AI involvement lowers overall artist perception and reduces behavioral engagement. As predicted, these effects are selective: they concentrate on attributions diagnostic of the artist's causal contribution while broad dispositional impressions remain unaffected. Audiences thus do not discount co-creation per se; they discount technological co-creation because it weakens the human basis of the attributions on which artist perception rests.
Removing the Human from Human Resources: Experimental Evidence on Applicant Reactions to AI-Mediated Recruitment
Co-authors: Kai Heinrich and Kim Siegling
AI-mediated recruitment promises fast, scalable hiring, but it also may alienate potential applicants who experience the technology without directly using or controlling it. Applicants are evaluated through AI-based systems and must infer how their data, identity, and opportunities are affected. Drawing on Value Sensitive Design and Contextual Integrity, we conceptualize applicants as affected non-users and argue that AI disclosures in job advertisements shape perceived fairness, recognition, and appropriate information flows. Across two vignette experiments with job advertisements, we examine how applicants respond to different levels of AI involvement and whether low-cost communication-based strategies can mitigate possible reactions. Study 1 shows that AI disclosure lowers willingness to apply and perceived fairness, especially for high-skilled job advertisements and independent of the level of AI involvement. Perceived fairness accounts for the effect on willingness to apply, and negative reactions are stronger among participants with lower technological affinity and more complex jobs. Study 2 shows that AI-based pre-screening increases concerns about being reduced to a number, while a video-based self-representation option and assurances that the AI scope is limited in the company do not mitigate these reactions. The findings highlight the need to design AI-mediated recruitment as a human-centered socio- technical process with meaningful oversight and contestability.
A Cashless Payment Option in Face-to-Face Fundraising
Co-authors: Adrian Chadi, Laszlo Goerke, and Konstantin Homolka
Cashless payments are ubiquitous, but if people are asked to contribute to charity face-to-face, they often prefer cash. A cashless payment option (CPO) may then even deter some individuals from donating. To investigate such behavioral phenomena, we carried out three separate experiments on the introduction of modern payment methods in face-to-face fundraising. We first conducted a field experiment in an art exhibition, where visitors could provide financial support after a guided tour. We randomly manipulated the donation opportunity, either offering a CPO or not. Additionally, we implemented two pay-whatyou-want (PWYW) treatment conditions (with and without CPO) by asking to compensate for the guided tour. Our results show that individuals hardly use the CPO in both the donation scenario and the PWYW scenario. Some types of visitors, such as frequent card users, were more likely to donate in cash because of the CPO. To explain this psychological phenomenon, we conducted a survey experiment on social pressure to give in fundraising, which shows that a CPO makes the excuse of not having cash available less credible. Moreover, our survey experiment shows that a CPO harms the awareness of need, which is a separate psychological driver of voluntary giving that could explain negative effects on donations. In line with this, a third experiment conducted at a Christmas market confirms that participants in face-to-face fundraising rarely use the CPO but react psychologically to its introduction by adjusting their willingness to donate.
Working Papers:
Alone at Home: The Impact of Social Distancing on Norm-Consistent Behavior. IWH Discussion Papers No. 08/2021. [link]
Co-author: Joschka Waibel
Around the globe, the Covid-19 pandemic has turned daily live upside down since social distancing is probably the most effective means of containing the virus until herd immunity is reached. Social norms have been shown to be an important determinant of social distancing behaviors. By conducting two experiments and using the priming method to manipulate social isolation recollections, we study whether social distancing has in turn affected norms of prosociality and norm compliance. The normative expectations of what behaviors others would approve or disapprove in our experimental setting did not change. Looking at actual behavior, however, we find that persistent social distancing indeed causes a decline in prosociality ―even after the relaxation of social distancing rules and in times of optimism. At the same time, our results contain some good news since subjects seem still to care for norms and become more prosocial once again after we draw their attention to the empirical norm of how others have previously behaved in a similar situation.
For a summary of our article, see our LSE COVID-19 blog post, or oekonomenstimme.org (in German).
Press coverage (all in German): Berliner Zeitung, n-tv.de. Further articles in Magdeburger Volksstimme (05.08.2021), Schweriner Volkszeitung (28.06.2021), Leipziger Zeitung and Hamburger Abendblatt (25.06.2021), Deutschlandfunk Nova (24.06.2021). Short report in „Deffner & Zschäpitz: Wirtschaftspodcast von WELT“ (29.06.2021).