Search

Community and organizational policies are typically designed in a top-down, centralized fashion, with limited input from impacted stakeholders. This can result in policies that are misaligned with community needs or perceived as illegitimate. How can we support more collaborative, participatory approaches to policy design? In this paper, we present PolicyCraft, a system that structures collaborative policy design through case-grounded deliberation. Building on past research that highlights the value of concrete cases in establishing common ground, PolicyCraft supports users in collaboratively proposing, critiquing, and revising policies through discussion and voting on cases. A field study across two university courses showed that students using PolicyCraft reached greater consensus and developed better-supported course policies, compared with those using a baseline system that did not scaffold their use of concrete cases. Reflecting on our findings, we discuss opportunities for future HCI systems to help groups more effectively bridge between abstract policies and concrete cases.

Emerging efforts in AI alignment seek to broaden participation in shaping model behavior by eliciting and integrating collective input into a policy for model finetuning. While pluralistic, these processes are often linear and do not allow participating stakeholders to confirm whether potential outcomes of their contributions are indeed consistent with their intentions. Design prototyping has long advocated for rapid iteration using tight feedback loops of ideation, experimentation, and evaluation to mitigate these issues. We thus propose policy prototyping for LLMs, a new process that draws inspiration from prototyping practices to enable stakeholders to collaboratively and interactively draft LLM policies. Through learnings from a real-world LLM policymaking initiative at an industrial AI lab, we motivate our approach and characterize policy prototyping with four guiding principles. Because policy prototyping emphasizes a contrasting set of priorities compared to previous approaches, we envision our approach to be a valuable addition to the methodological repertoire for pluralistic alignment.

Existing tools for laypeople to create personal classifiers often assume a motivated user working uninterrupted in a single, lengthy session. However, users tend to engage with social media casually, with many short sessions on an ongoing, daily basis. To make creating personal classifiers for content curation easier for such users, tools should support rapid initialization and iterative refinement. In this work, we compare three strategies -- (1) example labeling, (2) rule writing, and (3) large language model (LLM) prompting -- for end users to build personal content classifiers. From an experiment with 37 non-programmers tasked with creating personalized comment moderation filters, we found that with LLM prompting, participants reached 95\% of peak performance in 5 minutes, beating other strategies due to higher recall, but all strategies struggled with iterative refinement. Despite LLM prompting's better performance, participants preferred different strategies in different contexts and, even when prompting, provided examples or wrote rule-like prompts, suggesting hybrid approaches.

Large language models (LLMs) are increasingly capable of providing users with advice in a wide range of professional domains, including legal advice. However, relying on LLMs for legal queries raises concerns due to the significant expertise required and the potential real-world consequences of the advice. To explore \textit{when} and \textit{why} LLMs should or should not provide advice to users, we conducted workshops with 20 legal experts using methods inspired by case-based reasoning. The provided realistic queries ("cases") allowed experts to examine granular, situation-specific concerns and overarching technical and legal constraints, producing a concrete set of contextual considerations for LLM developers. By synthesizing the factors that impacted LLM response appropriateness, we present a 4-dimension framework: (1) User attributes and behaviors, (2) Nature of queries, (3) AI capabilities, and (4) Social impacts. We share experts' recommendations for LLM response strategies, which center around helping users identify `right questions to ask' and relevant information rather than providing definitive legal judgments. Our findings reveal novel legal considerations, such as unauthorized practice of law, confidentiality, and liability for inaccurate advice, that have been overlooked in the literature. The case-based deliberation method enabled us to elicit fine-grained, practice-informed insights that surpass those from de-contextualized surveys or speculative principles. These findings underscore the applicability of our method for translating domain-specific professional knowledge and practices into policies that can guide LLM behavior in a more responsible direction., Comment: 14 pages

Case studies commonly form the pedagogical backbone in law, ethics, and many other domains that face complex and ambiguous societal questions informed by human values. Similar complexities and ambiguities arise when we consider how AI should be aligned in practice: when faced with vast quantities of diverse (and sometimes conflicting) values from different individuals and communities, with whose values is AI to align, and how should AI do so? We propose a complementary approach to constitutional AI alignment, grounded in ideas from case-based reasoning (CBR), that focuses on the construction of policies through judgments on a set of cases. We present a process to assemble such a case repository by: 1) gathering a set of ``seed'' cases -- questions one may ask an AI system -- in a particular domain from discussions in online communities, 2) eliciting domain-specific key dimensions for cases through workshops with domain experts, 3) using LLMs to generate variations of cases not seen in the wild, and 4) engaging with the public to judge and improve cases. We then discuss how such a case repository could assist in AI alignment, both through directly acting as precedents to ground acceptable behaviors, and as a medium for individuals and communities to engage in moral reasoning around AI., Comment: MP2 workshop @ NeurIPS 2023

Communities and groups often need to make decisions based on social norms and preferences, such as when moderating content or building AI systems that reflect human values. The prevailing approach has been to first create high-level guidelines -- ``constitutions'' -- and then decide on new cases using the outlined criteria. However, social norms and preferences vary between groups, decision-makers can interpret guidelines inconsistently, and exceptional situations may be under-specified. In this work, we take inspiration from legal systems and introduce ``case law grounding'' (CLG), a novel workflow that uses past cases and decisions (\textbf{precedents}) to help ground future decisions, for both human and LLM-based decision-makers. We evaluate CLG against a constitution-only approach on two tasks for both types of decision-makers, and find that decisions produced with CLG were more accurately aligned to observed ground truth in all cases, producing a 3.3--23.3 \%-points improvement (across different tasks and groups) for humans and 9.2--30.0 \%-points (across different tasks and groups) for LLM agents. We also discuss other aspects where a case-based approach could augment existing ``constitutional'' approaches when it comes to aligning human and AI decisions.

Medical image segmentation modeling is a high-stakes task where understanding of uncertainty is crucial for addressing visual ambiguity. Prior work has developed segmentation models utilizing probabilistic or generative mechanisms to infer uncertainty from labels where annotators draw a singular boundary. However, as these annotations cannot represent an individual annotator's uncertainty, models trained on them produce uncertainty maps that are difficult to interpret. We propose a novel segmentation representation, Confidence Contours, which uses high- and low-confidence ``contours'' to capture uncertainty directly, and develop a novel annotation system for collecting contours. We conduct an evaluation on the Lung Image Dataset Consortium (LIDC) and a synthetic dataset. From an annotation study with 30 participants, results show that Confidence Contours provide high representative capacity without considerably higher annotator effort. We also find that general-purpose segmentation models can learn Confidence Contours at the same performance level as standard singular annotations. Finally, from interviews with 5 medical experts, we find that Confidence Contour maps are more interpretable than Bayesian maps due to representation of structural uncertainty., Comment: 10 pages content, 12 pages total. Accepted to HCOMP '23

To investigate the well-observed racial disparities in computer vision systems that analyze images of humans, researchers have turned to skin tone as more objective annotation than race metadata for fairness performance evaluations. However, the current state of skin tone annotation procedures is highly varied. For instance, researchers use a range of untested scales and skin tone categories, have unclear annotation procedures, and provide inadequate analyses of uncertainty. In addition, little attention is paid to the positionality of the humans involved in the annotation process--both designers and annotators alike--and the historical and sociological context of skin tone in the United States. Our work is the first to investigate the skin tone annotation process as a sociotechnical project. We surveyed recent skin tone annotation procedures and conducted annotation experiments to examine how subjective understandings of skin tone are embedded in skin tone annotation procedures. Our systematic literature review revealed the uninterrogated association between skin tone and race and the limited effort to analyze annotator uncertainty in current procedures for skin tone annotation in computer vision evaluation. Our experiments demonstrated that design decisions in the annotation procedure such as the order in which the skin tone scale is presented or additional context in the image (i.e., presence of a face) significantly affected the resulting inter-annotator agreement and individual uncertainty of skin tone annotations. We call for greater reflexivity in the design, analysis, and documentation of procedures for evaluation using skin tone., Comment: To appear in FAcct '23

When groups of people are tasked with making a judgment, the issue of uncertainty often arises. Existing methods to reduce uncertainty typically focus on iteratively improving specificity in the overall task instruction. However, uncertainty can arise from multiple sources, such as ambiguity of the item being judged due to limited context, or disagreements among the participants due to different perspectives and an under-specified task. A one-size-fits-all intervention may be ineffective if it is not targeted to the right source of uncertainty. In this paper we introduce a new workflow, Judgment Sieve, to reduce uncertainty in tasks involving group judgment in a targeted manner. By utilizing measurements that separate different sources of uncertainty during an initial round of judgment elicitation, we can then select a targeted intervention adding context or deliberation to most effectively reduce uncertainty on each item being judged. We test our approach on two tasks: rating word pair similarity and toxicity of online comments, showing that targeted interventions reduced uncertainty for the most uncertain cases. In the top 10% of cases, we saw an ambiguity reduction of 21.4% and 25.7%, and a disagreement reduction of 22.2% and 11.2% for the two tasks respectively. We also found through a simulation that our targeted approach reduced the average uncertainty scores for both sources of uncertainty as opposed to uniform approaches where reductions in average uncertainty from one source came with an increase for the other.

Online social platforms centered around content creators often allow comments on content, where creators moderate the comments they receive. As creators can face overwhelming numbers of comments, with some of them harassing or hateful, platforms typically provide tools such as word filters for creators to automate aspects of moderation. From needfinding interviews with 19 creators about how they use existing tools, we found that they struggled with writing good filters as well as organizing and revisiting their filters, due to the difficulty of determining what the filters actually catch. To address these issues, we present FilterBuddy, a system that supports creators in authoring new filters or building from existing filter lists, as well as organizing their filters and visualizing what comments are captured over time. We conducted an early-stage evaluation of FilterBuddy with YouTube creators, finding that participants see FilterBuddy not just as a moderation tool, but also a means to organize their comments to better understand their audiences., Comment: to be published in CHI Conference on Human Factors in Computing Systems (CHI '22)