The growing use of AI in public administration improves efficiency, yet its use in discretionary decisions raises concerns about fairness and legitimacy. While prior research examined decision stakes and Human–AI decision-making configurations separately, their combined effect on citizens’ perceptions of fairness and adoption remains underexplored. We conducted a mixed-method Wizard-of-Oz study (n=43) using an Intelligent-Self-Service-Kiosk. Participants completed a low-stakes (ID-renewal) and a high-stakes (social-housing) task under one of three decision-making configurations: AI alone, AI with human-supervision, and human with AI advice or recommendation. Quantitative analysis found no significant effects, highlighting the limits of standard metrics. However, qualitative interviews revealed that citizens valued human involvement, requiring meaningful over symbolic oversight. They emphasized interactive dialogue before decisions to capture their circumstances and after, to facilitate appeals. We contribute evidence of tensions between citizens’ desire for efficiency and need for human-control and fairness. We provide guidance for designing citizen-centered AI systems that align with democratic values.

In this paper, banking services, as the research object to explore the relationship among distributive fairness, procedural fairness, interactive fairness and informational fairness. The study found: informational fairness significant impact on the ...

Patching software theoretically leads to improvements including security critical changes, but it can also lead to new issues. For System Administrators (sysadmins) new issues can negatively impact operations at their organization. While mitigation options like test environments exist, little is known about their prevalence or how contextual factors like size of organization impact the practice of Patch Management. We surveyed 220 sysadmins engaged in Patch Management to investigate self-reported behaviors. We found that dedicated testing environments are not as prevalent as previously assumed. We also expand on known behaviours that sysadmins perform when facing a troublesome patch, such as employing a range of problem solving behaviours to inform their patching decisions.

Twitter accounts are public by default, but Twitter gives the option to create protected accounts, where only approved followers can see their tweets. The publicly visible information changes based on the account type and the visibility of tweets also depends solely on the poster's account type which can cause unintended disclosures especially when users interact. We surveyed 336 Twitter users to understand users' awareness of account information visibility, as well as the tweet visibility when users interact. We find that our participants are aware of the visibility of their profile information and individual tweets. However, the visibility of followed topics, lists, and interactions with protected accounts is confusing. Only 31% of the participants were aware that a reply by a public account to a protected account's tweet would be publicly visible. Surprisingly, having a protected account does not result in a better understanding of the account information or tweet visibility.

Static analysis tools (SATs) have the potential to assist developers in finding and fixing vulnerabilities in the early stages of software development, requiring them to be able to understand and act on tools’ notifications. To understand how helpful such SAT guidance is to developers, we ran an online experiment (N=132) where participants were shown four vulnerable code samples (SQL injection, hard-coded credentials, encryption, and logging sensitive data) along with SAT guidance, and asked to indicate the appropriate fix. Participants had a positive attitude towards both SAT notifications and particularly liked the example solutions and vulnerable code. Seeing SAT notifications also led to more detailed open-ended answers and slightly improved code correction answers. Still, most SAT (SpotBugs 67%, SonarQube 86%) and Control (96%) participants answered at least one code-correction question incorrectly. Prior software development experience, perceived vulnerability severity, and answer confidence all positively impacted answer accuracy.

In this study, we investigate the extent of polarisation among theist versus atheist groups on Arab Twitter and their networks. We find four main self-identified groups of Arab users that can be distinguished by different attitudes to religion. In addition to Atheists and Theists, there are Rationalists, who promote rational thinking regardless of religious affiliation, and Tanweeri, who promote religious reforms. Through a keyword search of Twitter account handles and biographies, we identified 2,673 active, public Twitter accounts that clearly state whether they are Atheists, Theists, Tanweeri or Rationalists and analysed the interactions among themselves and the accounts that are followed, retweeted, or mentioned the most in their networks. Depending on the network analysed, we found between four and seven sub-communities that highlight the rich socio-cultural context in which discussions of religion, non-religion, and religious reform unfold. While there was clear online polarisation between atheists and theists, Rationalist and Tanweeri accounts are spread among the two polarised sides, acting as natural bridges. We also found a clear separation between Arab atheists who engage with Arab accounts promoting atheism and those who primarily engage with Western accounts promoting atheism. We discuss implications for the study of religious debate and religious polarisation on social media.

Common anti-phishing advice tells users to mouse over links, look at the URL, and compare to the expected destination, implicitly assuming that they are able to read the URL. To test this assumption, we conducted a survey with 1929 participants recruited from the Amazon Mechanical Turk and Prolific Academic platforms. Participants were shown 23 URLs with various URL structures. For each URL, participants were asked via a multiple choice question where the URL would lead and how safe they feel clicking on it would be. Using latent class analysis, participants were stratified by self-reported technology use. Participants were strongly biased towards answering that the URL would lead to the website of the organization whose name appeared in the URL, regardless of its position in the URL structure. The group with the highest technology use was only minorly better at URL reading.

Applying regular patches is vital for the timely correction of security vulnerabilities, but installing patches also risks disrupting working systems by potentially introducing unknown errors. System administrators must manage the challenges of patching using a combination of reliance on best practice and available information to best match their organizations' needs. In this work, we study how patch-related activities are supported by the mailing list of the website PatchManagement.org which is dedicated to the task. We qualitatively coded 356 list emails sent between March and July, 2018, to understand how members interact with the list community. Based on our results, we argue that the mailing list is an example of an Online Community of Practice, where practitioners engage in communal learning and support. We find that the community supports members in multiple phases of the patching process by providing workarounds before a patch is available, guidance prioritizing released patches, and helping with post-patch trouble. Additionally, the community provides help around tool selection and facilitating discussions.

eHealth devices such as smart scales and wearable fitness trackers are a key part of many health technology solutions. However, these eHealth devices can be vulnerable to privacy and security related attacks. In this poster, we propose a security analysis framework for eHealth devices, called mH-PriSe, that will yield useful information for security analysts, vendors, health care providers, and consumers. We demonstrate our framework by analysing scales from 6 vendors. Our results show that while vendors strive to address security and privacy issues correctly, challenges remain in many cases. Only 5 out of 8 solutions can be recommended with some caveats whereas the remaining 3 solutions expose severe vulnerabilities.

Semantic fluency is a commonly used task in psychology that provides data about executive function and semantic memory. Performance on the task is affected by conditions ranging from depression to dementia. The task involves participants naming as many members of a given category (e.g. animals) as possible in sixty seconds. Most of the analyses reported in the literature only rely on word counts and transcribed data, and do not take into account the evidence of utterance planning present in the speech signal. Using data from Korean, we show how prosodic analyses can be combined with computational linguistic analyses of the words produced to provide further insights into the processes involved in producing fluency data. We compare our analyses to an established analysis method for semantic fluency data, manual determination of lexically coherent clusters of words.

OBJECTIVE: Cultural and health service obstacles affect the quality of pregnancy care that women from vulnerable populations receive. Using a participatory design approach, the Stress in Pregnancy: Improving Results with Interactive Technology group developed specifications for a suite of eHealth applications to improve the quality of perinatal mental health care. MATERIALS AND METHODS: We established a longitudinal participatory design group consisting of low-income women with a history of antenatal depression, their prenatal providers, mental health specialists, an app developer, and researchers. The group met 20 times over 24 months. Applications were designed using rapid prototyping. Meetings were documented using field notes. RESULTS AND DISCUSSION: The group achieved high levels of continuity and engagement. Three apps were developed by the group: an app to support high-risk women after discharge from hospital, a screening tool for depression, and a patient decision aid for supporting treatment choice. CONCLUSION: Longitudinal participatory design groups are a promising, highly feasible approach to developing technology for underserved populations.

Papers on voice interfaces for people with cognitive impairment or demenita only provide small snapshots of actual interactions, if at all. This is a major obstacle to the development of better interfaces. Transcripts of interactions between users and systems contain rich evidence of typical language patterns, indicate how users conceptualise their computer interlocutor, and highlight key design issues. In this paper, we introduce the CADENCE corpus and outline how it can be used to stimulate replicable research on inclusive voice interfaces. The CADENCE corpus is first data set of its kind to include rich data from people with cognitive impairment and free for research use. The corpus consists of transcribed spoken interactions between older people with and without cognitive impairment and a simulated Intelligent Cognitive Assistant and includes comprehensive data on users' cognitive abilities.

Reminders are often part of interventions to help older people adhere to complicated medication regimes. Computer-generated (synthetic) speech is ideal for tailoring reminders to different medication regimes. Since synthetic speech may be less intelligible than human speech, in particular under difficult listening conditions, we assessed how well older people can recall synthetic speech reminders for medications. 44 participants aged 50-80 with no cognitive impairment recalled reminders for one or four medications after a short distraction. We varied background noise, speech quality, and message design. Reminders were presented using a human voice and two synthetic voices. Data were analyzed using generalized linear mixed models. Reminder recall was satisfactory if reminders were restricted to one familiar medication, regardless of the voice used. Repeating medication names supported recall of lists of medications. We conclude that spoken reminders should build on familiar information and be integrated with other adherence support measures.

In this paper, we show how a large demographic data set that includes only high-level information about health and disability can be used to specify user requirements for people with specific needs and impairments. As a case study, we consider adapting spoken dialogue systems (SDS) to the needs of older adults. Such interfaces are becoming increasingly prevalent in telecare and home care, where they will often be used by older adults.

While there has been a lot of research on the usability of reminders and alarms in the work context, the home has been somewhat neglected despite the importance of reminder systems for telecare and assistive living systems. We conducted a comprehensive mixed-methods study into the requirements for useable and acceptable reminders in the home. The study consisted of a questionnaire (N=379), 6 focus groups, and 7 home tour interviews. Our results highlight the need for highly flexible and contextualized multimodal and multi-device reminder solutions that build on existing successful strategies for remembering in and around the home. We suggest that developers of home care reminder systems should design for diversity, context, priorities, autonomy, shared spaces, and optimal care.

In this paper we argue that Musicons, short samples from pieces of music are a useful way to present private but memorable reminder messages. We investigated accuracy, memorability and response times for short, medium, and long Musicons. User performance on the Musicons was also compared to short spoken reminders. The study consisted of two sessions a week apart. Quantitative measures were augmented with qualitative questions about associations and memories. Overall, participants achieved a high level of accuracy (89%) on the Musicons. Recognition was stable at 90% or better across sessions for users who were able to construct meaningful links between Musicons and the associated tasks. Optimal response times were achieved for medium-length 0.5 sec. Musicons. We conclude that Musicons are a viable option for alarms and notifications that combine the high learnability of Auditory Icons with the more private nature of Earcons.

When deciding whether to adapt relevant aspects of the system to the particular needs of older users, spoken dialogue systems often rely on automatic detection of chronological age. In this paper, we show that vocal ageing as measured by acoustic features is an unreliable indicator of the need for adaptation. Simple lexical features greatly improve the prediction of both relevant aspects of cognition and interactions style. Lexical features also boost age group prediction. We suggest that adaptation should be based on observed behaviour, not on chronological age, unless it is not feasible to build classifiers for relevant adaptation decisions.

We evaluated two strategies for alleviating working memory load for users of voice interfaces: presenting fewer options per turn and providing confirmations. Forty-eight users booked appointments using nine different dialogue systems, which varied in the number of options presented and the confirmation strategy used. Participants also performed four cognitive tests and rated the usability of each dialogue system on a standardised questionnaire. When systems presented more options per turn and avoided explicit confirmation subdialogues, both older and younger users booked appointments more quickly without compromising task success. Users with lower information processing speed were less likely to remember all relevant aspects of the appointment. Working memory span did not affect appointment recall. Older users were slightly less satisfied with the dialogue systems than younger users. We conclude that the number of options is less important than an accurate assessment of the actual cognitive demands of the task at hand.

Hearing loss as assessed by pure-tone audiometry (PTA) is sig-nificantly correlated with the intelligibility of synthetic speech. However, PTA is a subjective audiological measure that as-sesses the entire auditory pathway and does not discriminate between the different afferent and efferent contributions. In this paper, we focus on one particular aspect of hearing that has been shown to correlate with hearing loss: outer hair cell (OHC) function. One role of OHCs is to increase sensitivity and frequency selectivity. This function of OHCs can be assessed quickly and objectively through otoacoustic emissions (OAE) testing, which is little known outside the field of audiology. We find that OHC function affects the perception of human speech, but not that of synthetic speech. This has important implications not just for audiological and electrophysiological research, but also for adapting speech synthesis to ageing ears. Index Terms: speech synthesis, intelligibility, otoacoustic emissions, pure-tone audiometry

In order to synthesize the F0 contour of an utterance, we need an model of the F0 contour that both describes a complex contour by a small number of parameters and allows to generate all linguistically relevant pitch movements with the appropriate size and alignment. One possible solution is superpositional modelling. In this approach, F0 contours are described by a set of equations that correspond to longterm and short-term changes in the contour. A well-known

This is an acoustic study of prosodic effects on segmental durations with reference to syllable structure, stress, focus and tempo in Greek. Disyllabic nonsense words with one, or two, or three consonants in the initial syllable were examined in stressed/unstressed, focused/unfocused and normal/fast tempo productions in a carrier sentence contexts. The results indicate that: (1) syllabic onset branching has a bigger effect on the consonant than the vowel; (2) stress has a bigger effect on the vowel than the consonant; (3) focus effects are not substantially different from stress effects; and (4) tempo has a rather even effect on consonant and vowel durations.

A compositional account of the semantics of German prefix verbs in HPSG is outlined. We consider only those verbs that are formed by productive synchronic rules. Rules are fully productive if they apply to all base verbs which satisfy a common description. Prefixes can be polysemous and have separate, highly underspecified lexical entries. Adequate bases are determined via selection restrictions. 1 The Problem Determining the semantics of unknown words which can be derived from lexicon entries is highly desirable for natural language understanding (Light, 1996). In this paper, I sketch a compositional account of the semantics of German prefix verbs derived from a verbal base, concentrating on those verbs that can be generated by a productive word formation rule. Like (Witte, 1997), I assume that the meaning of most of these verbs can be derived compositionally by unifying the semantic representations of its constituents. Example: (1) durch + laufen ('through + to run') ) durchlaufe...

Control of prosody is essential for the synthesis of natural sounding speech. Text-to-speech systems tend to accent too many words when taking into account only the distinction between open-class and closed-class words. In the prominence-based approach [1], the degree of accentuation of a syllable is described in terms of a gradual prominence parameter. This paper presents the calculation of the prominence level of words based on their word class, the classes of the surrounding words, and their position in a clause. Rules predicting word prominence are derived from statistical analysis of a prosodic database. The hand-crafted rules are compared with the results of several machine learning algorithms on the same material. Furthermore, a perceptual test and an analysis of the resulting speech signals are carried out. 1. INTRODUCTION Good prosody control improves the naturalness of synthesised speech. Moreover, it aids comprehension. In practice, abstract prosodic labels are derived fro...