Philanthropy and Artificial Intelligence
Philanthropy has been largely absent from the debates on artificial intelligence (AI). This project investigates the relationship between AI and philanthropy from two perspectives. On the one hand, it explores how philanthropic organizations (POs) can play a crucial role in leading the Ethical and Inclusive Artificial Intelligence (EIAI) revolution from their privileged position as independent actors at the intersection of industry, government, and academia. Secondly, it addresses how the philanthropic sector can leverage AI to enhance governance and organizational development, data-driven decision-making, and the development of impact assessment strategies. The present project aims to address this gap by bringing together experts in the fields of data science, philanthropy, and ethics to generate solid academic and practice-informed contributions. These insights will provide a well-rounded reference point and guide POs’ in embracing and implementing AI, while encouraging POs active contribution to the promotion of EIAI practices.
This endeavor is a joint effort between the Chair in Behavioral Philanthropy and the Geneva Centre for Philanthropy (GCP) of the University of Geneva (UNIGE), supported by grants from Fondation Botnar and a strategic partner of the GCP which remains anonymous. The project benefits from the support of a diverse and multidisciplinary research team, including Dr. Lucia Gomez Teijeiro (postdoctoral researcher in Behavioral Philanthropy at UNIGE), Ms. Camilla della Giovampaola (Ph.D. student in International History and Politics at IHEID), Mr. Hubert Halope (Ph.D. student in management at UNIGE), Ms. Maria Cristiana Tudor (MSc in Neuroscience at UNIGE), and Ms. Nisa Thomas (BSc in computer sciences, UNIGE). In its initial phases, the project received contributions from Ms. Kamila Ciok, who designed the initial version of the survey on AI and Philanthropy, Dr. Joost Monks, who collaborated in the redaction of the concept paper summarising the key ideas of the project, and to fundraising efforts, Mr. Jayant Narayan who to commented the concept paper, and numerous experts in the fields of philanthropy and artificial intelligence who provided feedback during international conferences and workshops.
Stress and Honesty
This study examines the effects of acute stress on honesty and whether/how people’s protective values for honesty mediate this effect.
Honesty is vital for proper institutional and societal functioning, however, compliance with this norm often depends on trade-offs between acting morally and increasing personal gain. Previous experimental studies have demonstrated that individuals who hold honesty as a protected value are less inclined to give in to selfish temptations. These decisions frequently take place in highly stressful situations, yet we have very little knowledge on how stress impacts decisions to comply with honesty. Stress is known to alter behaviour by enhancing neural sensitivity to immediate rewards and decreasing activity in brain areas involved in response inhibition and goal selection. Thus, stress could affect honesty by: 1) making monetary rewards more attractive and 2) inducing habitual behaviour by reducing self-control. In the present study, we explore these potential mechanisms by acutely stressing participants and asking them to make decisions in three tasks measuring honesty. We posit that subjects’ protected values for honesty may act as a buffer against stress, mediating the relevance of the increased saliency of rewards on their choices. We will test four main hypotheses disentangling the specific ways in which individuals’ protected values for honesty, reward values, and self -control interact under acute stress. If the behavioral findings support our hypotheses, we will conduct fMRI studies to explore the brain areas involved.
Promoting Charitable Giving Through Tax Incentives
Donations by individuals play a vital part in supporting charitable organizations. To promote charitable giving, a common practice in several countries is to implement different subsidy schemes. Among these, two very common are a) rebate subsidies, in which governments (or other private donors) refunds a pre-established portion of the donation to the individual, and b) matching subsidies, in which a government (or other private donors) match the donation at a pre-announced rate. We conduct two preregistered experiments implementing public good game in which we investigate whether and how rebate, matching, or no subsidy schemes cause the reductions or increases of total, and net donations of earned money. We find that matching subsidy scheme results in more donations than rebate and no subsidy scheme. Interestingly, the rebate subsidy scheme does not result in more donations than no subsidy scheme, but does increase the cost of government above and beyond no subsidy. Moreover, we provide further evidence that there is a main effect of effort, and it's interaction with subsidies on total, and net donations of earned money.
Mapping the Swiss Philanthropic and Sustainable Finance Landscapes Through Big-Data
Switzerland occupies a starring position both in the philanthropic and the sustainable finance global ecosystems, counting with 10755 currently active philanthropic organizations (POs) and 231 companies listed sin the Swiss Stock Exchange. Applying cutting-edge machine learning techniques on detailed annotated and updated databases, we aim to understand how diverse those sectors are in terms of their missions, motivations or sustainability goals and how interconnected they are in terms of shared interests in those domains. This project seeks to provide open source resources directed to ultimately promote synergies between the different actors represented on the highly complex Swiss ecosystem.
This project will investigate whether/why people choose to expose themselves to negative tail risk in circumstances where one major loss can eliminate all previous gains.
Under negative tail risk, or the risk of loss due to a rare event as predicted by a probability distribution, one ruinous loss can wipe out a previous streak of small wins. However, rather than avoiding these situations, a large body of evidence shows that many people, from novices to experts, will continuously expose themselves to negative tail risk for small rewards until the colossal loss eventually occurs, a behavior dubbed the picking pennies bias. Explanations for this behavior may be cognitive (i.e. people have a limited understanding of the risk they are taking) or motivational (i.e. the utility of gambling theory), with previous findings indicating the involvement of motivational factors after cognitive factors were controlled for. When curbing motivational factors, self-control was shown to be a key element, as when given the opportunity, people often made use of commitment devices to keep themselves from picking pennies. In collaboration with Payzan-LeNestour, the developer of the original experiment, our project will expand on previous findings by including conditions with high/low rewards and losses, and collecting physiological data to explore how arousal might correlate with risk-taking in these settings. Once these findings are successfully replicated, we will proceed to an fMRI study to identify the brain areas underlying this bias.
Brand familiarity is an important and frequently used concept in marketing research and practice. Existing measures of brand familiarity typically rely on subjective self-reports and Likert scales. Here we develop and empirically test two implicit measures to quantify brand familiarity. Based on research in visual attention and computer image processing, observers in a first visual search task are incentivized to quickly find a target brand among varying numbers of competitor brands. In the second approach, we measure the speed at which observers can identify a target brand that is gradually revealed. Both approaches are validated in four preregistered experiments. Results show that reaction times predict brand familiarity on an individual level above and beyond conventional self-reports, even when controlling for “bottom-up” visual features of the brand logo. Our findings offer an innovative way to objectively measure brand familiarity and they contribute to the understanding of consumer attention.
Virtual Reality and Donations
In collaboration with the International Red Cross Committee (ICRC), this project will explore whether virtual reality is an efficient and feasible tool for eliciting donations.
The media richness theory examines a medium’s ability to facilitate understanding, positing that richer media, which include higher psychological perceptions of other people, reduce ambiguity and equivocality. However, richer media is not always preferable. Research exploring new media has found a ‘paradox of richness’, where a high degree of social presence increases performance on simple/well-known tasks, but decreases it on more complex tasks. Thus, when the aim is not to teach specific details, but to induce or enhance a more general response to an object, such as increased social responsibility, prosocial attitudes, and the intent to donate, it may be that media high in richness, such as virtual reality (VR), which increase social presence, immersion, and empathy, could be ideal. Indeed, although it is a relatively young field, several studies have found that, compared to videos, VR increases empathy, social responsibility, and the intention to donate/actual donations. In addition, social presence was identified as a key mediator of donation intention. This study will examine whether VR scenarios increase donations to the ICRC in comparison to videos. Our hypothesis is that the VR condition will elicit higher levels of empathy and social presence, and increase donations compared to the video condition. Moreover, by including a pre-task in which participants can earn money, we will be able to measure donations themselves, rather than merely the intention to donate. As well, we aim to explore what elements of VR drive these differences, namely a sense of presence, empathy, agency, and the participants’ role in the scenario (person experiencing a crisis/beneficiary, aid worker). In addition to behavioral and self-report measures, we will also collect physiological data (electrodermal and heart rate activity). Our hypothesis is that VR will lead to increased physiological arousal and we will probe how these measures correlate with subjective measures and behavioral results.