Helping Visually Impaired People Take Better Quality Pictures

Perception-based image analysis technologies can be used to help visually impaired people take better quality pictures by providing automated guidance, thereby empowering them to interact more confidently on social media. The photographs taken by visually impaired users often suffer from one or both of two kinds of quality issues: technical quality (distortions), and semantic quality, such as framing and aesthetic composition. Here we develop tools to help them minimize occurrences of common technical distortions, such as blur, poor exposure, and noise. We do not address the complementary problems of semantic quality, leaving that aspect for future work. The problem of assessing, and providing actionable feedback on the technical quality of pictures captured by visually impaired users is hard enough, owing to the severe, commingled distortions that often occur. To advance progress on the problem of analyzing and measuring the technical quality of visually impaired user-generated content (VI-UGC), we built a very large and unique subjective image quality and distortion dataset. This new perceptual resource, which we call the LIVE-Meta VI-UGC Database, contains 40K real-world distorted VI-UGC images and 40K patches, on which we recorded 2.7M human perceptual quality judgments and 2.7M distortion labels. Using this psychometric resource we also created an automatic blind picture quality and distortion predictor that learns local-to-global spatial quality relationships, achieving state-of-the-art prediction performance on VI-UGC pictures, significantly outperforming existing picture quality models on this unique class of distorted picture data. We also created a prototype feedback system that helps to guide users to mitigate quality issues and take better quality pictures, by creating a multi-task learning framework.

A Perspective on Building Ethical Datasets for Children's Conversational Agents

Artificial intelligence (AI)-powered technologies are becoming an integral part of youth's environments, impacting how they socialize and learn. Children (12 years of age and younger) often interact with AI through conversational agents (e.g., Siri and Alexa) that they speak with to receive information about the world. Conversational agents can mimic human social interactions, and it is important to develop socially intelligent agents appropriate for younger populations. Yet it is often unclear what data are curated to power many of these systems. This article applies a sociocultural developmental approach to examine child-centric intelligent conversational agents, including an overview of how children's development influences their social learning in the world and how that relates to AI. Examples are presented that reflect potential data types available for training AI models to generate children's conversational agents' speech. The ethical implications for building different datasets and training models using them are discussed as well as future directions for the use of social AI-driven technology for children.

CHALLENGES FOR BLIND AND VISION IMPAIRED USERS OF A VISUAL QUESTION ANSWERING TOOL: IMPLICATIONS FOR AGING ADULTS

Recent technological developments provide individuals with vision impairment the transformative ability to upload pictures they take and promptly receive descriptions from remote workers. This study aimed to: identify challenges for visually impaired individuals to use such technology to obtain health-related information and provide recommendations for crowd-workers and the future development of assistive artificial intelligence (AI) design. In spring and summer of 2019, we analyzed 265 images of medication packages submitted by users of a visual question answering (VQA) application called VizWiz -- a smartphone application that provides near realtime assistance to visually impaired users by employing crowd-workers. We developed a 4-category coding scheme to analyze image quality, with two independent coders achieving excellent intercoder reliability (85%-95%). Of the 265 images, we found less than half were legible (46%), contained clear indicators for information sought (40%), or had minimum background noise (40%); while only a small percentage contained complete information (6%). Through thematic analysis of the data, we also highlight seven challenges with queries submitted by vision impaired users . Based on our findings, we make recommendations for the future design of VQA technologies, such as VizWiz, for visually impaired users. We also suggest that there is both great need and potential for user-centered design research to significantly enhance such assistive technologies. While this study did not focus exclusively on data submitted by aging adults, many VizWiz users are, in fact, aging adults, and such assistive technologies have strong implications for the design of assistive technology for this age group.

Involvement of receptor-bound protein methyltransferase PRMT1 in antiviral and antiproliferative effects of type I interferons

Protein arginine N-methyltransferase (PRMT1) is one of the proteins that bind to the intracytoplasmatic domain of the IFNAR-1 chain of the type I interferon (IFN) receptor system. The attachment is specific and is not seen with PRMT2, another member of this protein family. Antisense PRMT1 cDNA constructs expressed under the early cytomegalovirus (CMV) promoter were transfected into HeLa cells, and stable transformants were selected. Antibodies to PRMT1 were used to identify clones with reduced PRMT1 expression. In such clones, IFN-beta inhibited three to five times less the replication of vesicular stomatitis virus (VSV) than in the original HeLa cells. The antiproliferative effect of IFN-beta was also reduced up to fivefold in the clones with low PRMT1 expression. No difference was seen when IFN-gamma was used alone to inhibit cell growth. The protein methylating enzyme, bound to IFNAR-1, appears to regulate positively the biologic activity of type I IFN.