As our awareness of autism and ableism continues to increase, so does our understanding of ableist language towards autistic people. Such language poses a significant challenge in NLP research due to its subtle and context-dependent nature. Yet, detecting anti-autistic ableist language remains underexplored, with existing NLP tools often failing to capture its nuanced expressions. We present AUTALIC, the first dataset dedicated to the detection of anti-autistic ableist language in context, addressing a significant gap in the field. AUTALIC comprises 2,400 autism-related sentences collected from Reddit, accompanied by surrounding context, and annotated by trained experts with backgrounds in neurodiversity. Our comprehensive evaluation reveals that current language models, including state-of-the-art LLMs, struggle to reliably identify anti-autistic ableism and diverge from human judgments, underscoring their limitations in this domain. We publicly release our dataset along with the individual annotations, providing an essential resource for developing more inclusive and context-aware NLP systems that better reflect diverse perspectives.

Annotating text for subjective tasks, such as labeling ableist and anti-autistic texts, is a challenge that has attracted significant attention as commonly adopted annotation paradigms, e.g., using majority voting, fall short in capturing the nuances of hate speech or bias annotations. Labeling ableist and anti-autistic texts presents similar challenges in addition to the need for familiarity with autism and anti-autistic discrimination. In this paper, we adopt a collaborative and annotator-centric approach to study the impact of various annotation techniques. We recruit 6 participants to annotate sets of sentences from our 11,596 sentence corpus. The groups annotate through schemes focused on score-based classification, algorithmic labeling, and comparison-based labeling to identify instances of anti-autistic ableist speech. As a result of changes in annotation schemes, our annotator groups shift from a worse-than-chance agreement to moderate agreement. This suggests that implementing annotator group discussion and collecting annotator feedback is likely to result in improved agreement scores in difficult and highly subjective tasks. Our results highlight the importance of a collaborative approach in highly subjective classification tasks as it may lead to an improved understanding of their own biases, and large improvements in agreement scores, particularly among annotators with higher rates of disagreement. Warning: This paper contains examples that may be offensive or upsetting, including explicit slurs used against people with disabilities.

The marginalization of autistic people in our society today is multi-faceted as it includes violence that is both physical and ideological in nature. It is rooted in the dehumanization, infantilization, and masculinization of autistic people and pervasive even in contemporary research studies that continue to echo ableist ideologies from the past. In this work, we identify how HRI research reproduces systemic social inequalities and explain how they align with historical misrepresentations, and other systemic barriers. We analyzed 142 papers focusing on HRI and autism published between 2016 and 2022. We critique these studies through a mixed-methods analysis of their definition of autism, study designs, participant recruitment, and results. Our findings indicate that HRI research stigmatizes autism in three dimensions - 1) the pathologization of autism, 2) gender and age-based essentialism, and 3) power imbalances. Our work uncovered that about 90% of HRI research during the timeline explored excluded the perspectives of autistic people, particularly those from understudied groups. We recommend broadening the inclusion of autistic people, considering research objectives beyond clinical use, and diversifying collaborations, foundational works considered, & participant demographics for more inclusive future work.

The availability of mobile money in Kenya has positively impacted commerce, financial transaction efforts, and the ability of individuals to receive and save their money. The addition of mobile loan applications provides access to loans without the hassle of going to a physical bank and, in some cases, completing paperwork. While it has its benefits, limited protection of user data has been a cause for concern. User complaints prompted a change in the mobile loan industry, requiring applications to be licensed and banning the use of specific permissions for Android versions of the apps placed in the Google Play Store. We investigate the impact of this change and explore ways to improve regulation by reviewing 30 licensed (n=15) and unlicensed (n=15) Kenyan-targeted digital lender apps. The results suggest that regulation has not yet had a significant impact on digital lender app development and thus encourages government-supported development guidelines and audits.

Due to greater accessibility, diverse game options, and the social experiences provided by gaming platforms, the number of children engaging with the Roblox platform has increased over time. This increase in gaming from children of all ages has led to the parental challenge of balancing child safety with fun. Although parents typically know when their children are online, it can be challenging to trust that Roblox will protect data and minimize risky experiences due to the publicized criticisms of inadequate protections and the requirement of parent engagement. In this study, we examine parental Roblox concerns by (1) reviewing the game’s privacy policy and features, (2) characterizing parental concerns expressed on Reddit, and (3) surveying adults about their Roblox opinions. Our findings indicate gaps exist between what safety features Roblox provides and what parents need. Additionally, Roblox could improve how they convey their privacy and security practices to players and parents.

Human-AI teams are increasingly prevalent in various domains. We investigate how the decision-making of a team member in a human-AI team impacts the outcome of the collaboration and perceived team-efficacy. In a large scale study on Mechanical Turk (n=125), we find significant differences across different decision making styles and disclosed AI identity disclosure in an AI-driven collaborative game. We find that autocratic decision-making negatively impacts team-efficacy in Human-AI teams, similar to its effects on human-only teams. We find that decision making style and AI-identity disclosure impacts how individuals make decisions in a collaborative context. We discuss our findings of the differences of collaborative behavior in human-human-AI teams and human-AI-AI teams.

SMS-based two-factor authentication (2FA) is the most widely deployed 2FA mechanism, despite the fact that SMS messages are known to be vulnerable to rerouting attacks, and despite the availability of alternatives that may be more secure. This is for two reasons. First, it is very effective in practice, as evidenced by reports from Google and Microsoft. Second, users prefer SMS over alternatives, because text messaging is already part of their daily communication. Accepting this practical reality, we developed a new SMS-based protocol that makes rerouting attacks useless to adversaries who aim to take over user accounts. Our protocol delivers one-time passwords (OTP) via text message in a manner that adds minimal overhead (to both the user and the server) over existing SMS-based methods, and is implemented with only small changes to the stock text-message applications that already ship on mobile phones. The security of our protocol rests upon a provably secure authenticated key exchange protocol that, crucially, does not place significant new burdens upon the user. Indeed, we carry out a user study that demonstrates no statistically significant difference between traditional SMS and our protocol, in terms of usability.

Disinformation is false information created to mis-lead public belief. In politics, disinformation is often used to persuade public opinion and achieve political victory. False information can be spread through word of mouth, news articles, images, and videos. The use of manipulated media is becoming a source of concern for the general public. Researchers are exploring the impact of manipulated media on public opinion. While prior research has investigated the effects of various news articles and images, to our knowledge, researchers have yet to explore the impact of manipulated videos of political figures on public opinion. We conducted a between-subjects user study with 420 participants. First, participants viewed edited or unedited videos related to a political candidate and then answered some survey questions regarding their opinion on the political figures. Our findings suggest that video headlines and news sources have influenced user interpretations and beliefs. However, many participants were skeptical about the authenticity of the videos. Overall, we see that a viewer’s political leanings do affect their judgment of a video’s authenticity but the Figure involved and the context of the video matter as well.

With the growth of technology that produces misinformation, there is a growing need to help users identify emerging types of fake media such as edited images and manipulated videos. In this work, we conduct a mixed-methods investigation into how we can provide provenance indicators to assist users in detecting newer forms of fake media. Specifically, we interview users regarding their experiences with different misinformation modes (text, image, video) to inform the design and content of indicators for previously unexplored media, especially fake videos. We find that media provenance – the source of the information – is a key heuristic used to evaluate all forms of fake media, and a heuristic that can be addressed by emerging technology. Thus, we subsequently design and investigate the use of provenance indicators to help users identify fake videos. We conduct a participatory design study to develop and design provenance indicators and evaluate participant-designed indicators via both expert evaluations and quantitative surveys (n=1,456) with end-users. Our results provide concrete design guidelines for the emerging issue of fake media. Our findings also raise concerns regarding users’ tendency to overgeneralize indicators used to assist users in identifying misinformation, suggesting the need for further research on warning design in the ongoing fight against misinformation.

Robocalls interrupt daily activity, cause financial harm, and influence users to ignore calls from unfamiliar numbers. Service providers and developers have created Anti-Robocall applications to attempt to restore trust in the phone and decrease the impact of robocalls on daily life. However, whether or not such applications meet accessibility standards and are therefore usable by vulnerable populations, particularly the visually impaired, is unknown. In this paper, we use a combination of the W3C's Mobile Web Content Accessibility Guidelines (MWCAG) and interviews with 11 visually impaired users to establish accessibility metrics for Anti-Robocall applications. We then evaluate 56 Anti-Robocall applications for Android to assess whether they met the needs of the visually impaired community. Our results indicate that 100% of the applications fail to meet all basic accessibility guidelines including minimum color contrast, button labels (to assist screen readers), and automatic audible alerts. As a result, we show that despite the availability of a variety of tools to help developers identify and correct these problems, this important class of applications does not meet basic accessibility requirements. We conclude by suggesting viable paths forward that ensure inclusion and protection for the visually impaired community.

Credit and debit cards enable financial transactions at unattended "pay-at-the-pump" gas station terminals across North America. Attackers discreetly open these pumps and install skimmers, which copy sensitive card data. While EMV (“chip-and-PIN”) has made substantial inroads in traditional retailers, such systems have virtually no deployment at pay-at-the-pump terminals due to dramatically higher costs and logistical/regulatory constraints, leaving consumers vulnerable in these contexts. In an effort to improve security, station owners have deployed security indicators such as low-cost tamper-evident seals, and technologists have developed skimmer detection apps for mobile phones. Not only do these solutions put the onus on consumers to notice and react to security concerns at the pump, but the efficacy of these solutions has not been measured. In this paper, we evaluate the indicators available to consumers to detect skimmers. We perform a comprehensive teardown of all known skimmer detection apps for iOS and Android devices, and then conduct a forensic analysis of real-world gas pump skimmer hardware recovered by multiple law enforcement agencies. Finally, we analyze anti-skimmer mechanisms deployed by pump owners/operators, and augment this investigation with an analysis of skimmer reports and accompanying security measures collected by the Florida Department of Agriculture and Consumer Services over four years, making this the most comprehensive long-term study of such devices. Our results show that common gas pump security indicators are not only ineffective at empowering consumers to detect tampering, but may be providing a false sense of security. Accordingly, stronger, reliable, inexpensive measures must be developed to protect consumers and merchants from fraud.

Access to credit can provide capital crucial to both businesses and individuals. Unfortunately, for large parts of the developing world, access to credit is not available because customers often lack the traditional data used by lenders to make such decisions (e.g., verifiable payroll statements, property ownership documents). Emerging online credit services address this need through the use of non-traditional creditworthiness data, which many believe to include user geolocation and social network information. While such systems both potentially expand credit availability and improve usability through instant evaluation, their security and privacy practices remain opaque. In this paper, we perform the first comprehensive security analysis of the emerging online credit space. To provide improved transparency, we select 51 representative companies across the industry, analyze their privacy policies and compare them to the sensitive data types mobile applications actually gather. We then evaluate the configuration of connections between mobile apps and their supporting servers to determine whether they securely handle such data. Our analysis demonstrates significant security and privacy issues across this burgeoning industry, including the gathering of previously undisclosed data types and widespread mis-configuration of encryption. We conclude by discussing our efforts to work with partners in and around the industry to improve these issues.

Fully autonomous or “self-driving” vehicles are an emerging technology that may hold tremendous mobility potential for individuals who are visually impaired who have been previously disadvantaged by an inability to operate conventional motor vehicles. Prior studies however, have suggested that these consumers have significant concerns regarding the accessibility of this technology and their ability to effectively interact with it. We present the results of a quasi-naturalistic study, conducted on public roads with 20 visually impaired users, designed to test a self-driving vehicle human–machine interface. This prototype system, ATLAS, was designed in participatory workshops in collaboration with visually impaired persons with the intent of satisfying the experiential needs of blind and low vision users. Our results show that following interaction with the prototype, participants expressed an increased trust in self-driving vehicle technology, an increased belief in its likely usability, an increased desire to purchase it and a reduced fear of operational failures. These findings suggest that interaction with even a simulated self-driving vehicle may be sufficient to ameliorate feelings of distrust regarding the technology and that existing technologies, properly combined, are promising solutions in addressing the experiential needs of visually impaired persons in similar contexts.

In 2012, about one-third of voters with disabilities reported having issues when voting in a polling place. Although the Help America Vote Act (HAVA) was passed in 2002, it is clear that there is room for improvement within the domain of accessible voting. Prime III is a voting technology that addresses many issues that plague other accessible voting systems. By addressing the needs of different communities, Prime III has become a ballot marking system that allows all voters to vote on one machine. This demonstration will showcase the accessibility features of Prime III and how it can be used in elections.