Background

Scope of the half-day workshop at AVI2020 on Ischia Island (Naples)

Artificial Intelligence (AI) is a trending topic in various application domains. Popular application areas reach from health care [1][2], trading [3], driverless cars, humanoid robots [2] to people's everyday life where shopping websites recommend products, video streaming providers such as Netflix suitably recommend films and social media platforms individually decide upon content relevancy for each user [3]. Although AI is a trending topic, it is a topic that has fascinated humans since ancient times [2]. Examples are the Greek Mythology where statues were brought to life or Marry Shelly's Dr. Victor Frankenstein, a man-made creature that acquires human behavior [2]. The modern computer-based AI are closely linked to the 1950s and Alan Turing's question regarding thinking capabilities of machines [4]. Important milestones were the introduction of Artifical Neural Networks by Bernard Widrow in the 1950s [5] and the success of both, IBM's Deep Blue chess AI [2] as well as Alphabet's AlphaGo Zero AI for the asian game of Go [4] defeating the at the time world champions for the respective game. The latter AI learned its skills in a self-contained manner by playing against itself with neither human instructions nor historical data [4]. AI and AI-relevant terminology is standardized in ISO/IEC JTC 1/SC 42 standard [6] which declares AI to be a form of intelligence displayed by machines other than natural intelligence which is displayed by humans and animals [6]

The term and application area of Big Data is closely connected to AI [6] since big data provides AI with the possibility of deriving, validating, applying and enhancing statistical models through computation of huge amounts of data, whereas AI and its statistical approach of Machine Learning (ML) provides algorithms to exploit to exploit Big Data and its potential. A definition for big data is provided by Doug Laney [7] who describes data to be big data once the three v's volume (big amount of data), velocity (high frequent data inflow) and variety (ambiguous data manifestations regarding e.g. data format, data structure or data semantics) [8] are met. Although AI models and algorithms support the analysis of big data, human interaction and perception by different user stereotypes (from domain experts such as data engineers to management-level end user stereotypes) are the pivotal elements for exploiting big data [7]. Starting with an AVI Road Mapping Workshop in 2016, Bornschlegl et al. introduced in [9] a theoretical reference model called Information Visualization for Big Data (IVIS4BigData) that close[s] the gap in research with regard to Information Visualization challenges of Big Data analysis as well as context awareness [7]. IVIS4BigData covers the whole lifecycle of Big Data analysis from Raw Data Source into Views that enable the different user stereotypes participating in the process to gain Insight [9]. Furthermore, the reference model considers the for this purpose necessary advanced visual user interfaces as well as the underlying data storage, computation and service infrastructure to be decisive [9]. This workshop aims to carry on the comprehensive research and results of IVIS4BigData since 2016.

The utilization of AI for performing data transformation (analysis) in big data is well-established and a fundamental element of the IVIS4BigData reference model [7]. Nevertheless, further applications beyond analysis are rarely discussed in scientific publications although existing problems and challenges emphasize to do so: Although multiple user stereotypes with different technical backgrounds and skills participate in and benefit from the process of Big Data analysis, the valuable input of the Data Scientist user stereotype is still indispensable with today's analysis system implementations. As this user stereotype is still a relatively new profession and in comparison to the job market's demand very rare, especially small research facilities and Small and Medium-sized Enterprises (SMEs) will be confronted with recruitment challenges [7]. Since Big Data analysis involves multiple user stereotypes with different technical backgrounds and skills, advanced visual user interfaces for big data analysis are required to distinct between their different skill level. This raises a challenge for more individualization of the user interfaces' capabilities. Especially since comprehension and perception of different users varies drastically: Although a user interface might theoretically empower all users to perform a necessary interaction, there exists a subset of users that do not identify the right interface to do so. In addition, the variable nature of big data (variety of data manifestations with a high velocity regarding data inflow [10]) urges that there exists the further user interface challenge: a user interface that is perfectly designed for a certain task and a certain user competence profile doesn't suit the exact same user for the same task when the underlying artifact (the data) altered. Even if storage and computing resources' prices decline, the rising amounts of data [7] research institutions as well as private companies are confronted with, generate significant costs and raise a monetary challenge. The application of AI to solve or mitigate these problems raises further challenges regarding the comprehension of AI's reasoning and decision making for human users (lack of explainability and transparency cause difficulties regarding human-comprehensibility, debugging, fairness and accountability [4]). In addition, dynamic and evolving AI models cause user irritation due to altering system behavior (lack of reproducability and reliability. More AI application domains and growing AI-awareness of society raise further challenges regarding regulatory and legal requirements that need to be taken into account for designing AI based systems. This proposal will address these issues with a special focus on supporting intelligent advanced visual user interfaces for Big Data Analysis.

The purpose of this research road-mapping workshop is threefold: Firstly, we aim to consolidate information, technical details, and research directions from the diverse range of academic and industrial R&D projects currently available. Secondly, based on visions of future infrastructures, identify gaps in the current state of the art and thirdly, to identify new areas of research which require funding and support along with new areas for collaboration outside.

• The aim of this workshop is to bring together researchers active in these areas to produce a research road map for AI supporting Advanced Visual Big Data Analysis with a special focus on intelligent advanced visual interfaces that addresses needs not otherwise well met. This road map will identify current research and development and will identify potential gaps in knowledge and execution.
• We articulate a clear research, education and training agenda that tracks and builds upon existing and emerging technologies and that respond to future visions of AI and Advanced Visual Interface infrastructures supporting Advanced Visual Big Data Analysis. The process will develop a baseline by drawing upon all of our previous workshops and other activities within our recent research projects and will produce a research, technology and development road map. During the workshop we will map out future visions and complete a gap analysis. By treating this over the course of the day we can feedback to refine the baseline (to be captured into a workshop report and corresponding publications) and real research, technology and development gaps, which should be detailed in the road map. For this workshop we are adopting idealization, knowledge capture and road mapping approaches and activities developed during the H2020 projects EDISON, SENSECARE and METAPLAT.
• In this way, the overall scope and goal of the workshop is to achieve a road map, which can support the acceleration in research, education and training activities by means of transforming, enriching, and deploying AI models and algorithms as well as intelligent advanced visual user interfaces supporting creation, configuration, management and usage of distributed Big Data Analysis. Thereby, the research, education and training road map will pave the way towards establishing of an infrastructure for an AI-enabled visual user interface tool suite supporting Big Data Analysis and corresponding research activities sharing distributed research resources (i.e., data, tools, and services) by adopting common existing open standards for ML techniques, access, analysis and visualization, realizing a ubiquitous collaborative workspace for researchers which is able to facilitate the research process and its Big Data Analysis applications.

Researchers are welcome that work on different scopes of these problem spaces.