Driven by methodological concerns, theoretical considerations, and previous evidence, I systematically test the validity of common dictator game variants with probabilistic payoffs. Using a unified experimental framework, I include four approaches and compare them to a standard dictator game: involving fewer receivers than dictators, paying only some players, paying only some decisions, and role uncertainty. I also relate transfers in the dictator game variants to established complementary individual difference measures of prosociality: social value orientation, personal values, a donation to charity, and the Big Five personality factor agreeableness. My data shows that the standard dictator game presents the expected correlations with the complementary measures of prosociality. Involving fewer receivers yields comparably valid results. By contrast, when only some players or decisions are paid or, particularly, when subjects face role uncertainty, the expected associations with complementary prosociality measures are distorted. Under role uncertainty, generosity is also significantly biased upward. I conclude that the validity of dictator game outcomes is highly sensitive to the applied methods. Not all dictator game variants can be recommended for the valid measurement of social preferences.

在方法论上的关注，理论上的考虑和先前的证据的驱使下，我系统地测试了带有概率收益的普通独裁者博弈变量的有效性。使用统一的实验框架，我包括四种方法，并将它们与标准的独裁者游戏进行比较：接收者少于独裁者，只付给一些球员，只付一些决定，以及角色不确定性。我还将独裁者游戏变体中的转移与已建立的互补性个体差异度量联系起来：社会价值取向，个人价值，对慈善机构的捐赠以及大五人格因素的可喜性。我的数据表明，标准独裁者博弈与亲社会性的互补度量呈现了预期的相关性。让更少的接收者参与进来可以获得相当有效的结果。相比之下，当仅支付一些参与者或决定时，或者特别是当主体面临角色不确定性时，预期的与补充性亲权措施的关联就会失真。在角色不确定的情况下，慷慨大方也明显偏向上方。我得出结论，独裁者博弈结果的有效性对所采用的方法高度敏感。并非所有独裁者游戏变体都可以推荐用于有效衡量社会偏好。