Today's business environments are characterized by an indisputable growth in the volume, complexity and multivariate nature of business processes, data structures and sources. For the business end-users, this is many times an overwhelming and demotivating experience when interacting with rich business data visualizations and scenarios. As they need to explore demanding use cases, create fast an understanding and make informed decisions so to meet their business goals. This position paper addresses this challenge by introducing a human-centred model that consists of four main dimensions: User, Visualizations, Data, and Tasks, and which is maintained at the core of an adaptive data analytics platform in the business domain. The aim is two-fold: To provide (a) best-fit representation of data for the unique end-users, and (b) personalized and transparent path of exploration towards accomplishing purposeful end-to-end business activities. Thus, enabling explainable and intuitive interactions for accurate decision making and problem solving - saving time and costs.
1. Introduction ing complex patterns. However, according to IBM, every day we create 2.5 quintillion bytes Modern business intelligence and data ana- of data – so much data that 90% of all the lytics platforms use real time visual analytics data in the world today has been created in to continuously monitor and analyze business the last two years alone [1]. These data come transactions and historical data so to facil- from a variety of sources and in diverse foritate real-time decision support. The result mats, both structured and unstructured, creof this process is then exported into various ating a business ecosystem that brings new standard format artifacts (e.g., tabular forms, insights but also generates a number of comgraphs, etc.) ofering customization options plications and problems (e.g., delays in realto end-users as means for (visual) data ex- time processing, inefective delivery of multiploration for obtaining insights and unveil- purpose information). As a consequence, this Joint Proceedings of the ACM IUI 2021 Workshops, April may disorient end-users that need to navi13–17, 2021, College Station, USA gate and take decisions faster than ever when " camyrotos@uclan.ac.uk (C. Amyrotos); performing their daily business activities uspgandreou@uclan.ac.uk (P. Andreou); ing data analytic solutions. Although such p~anhatgtpio:/t/issc.greart.mcsa.uncayk.oasc@.cys/appg.ceormma(nP.(GP.eGrmeramnaaknoask)os) platforms may provide data visualizations that are considered more usable than others [2], often their recipients (i.e., the decision makers) are overloaded from the vast amount of © 2021 Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution CPWrEooUrckReshdoinpgs IhStpN:/c1e6u1r3-w-0s.o7r3g 4(C.C0EIEnUtUeRrnRat-iWonaSlW(.CooCrrBgkY)s4.h0)o.p Proceedings visual information, which in turn severely de- 2. Background creases their ability to eficiently assess situ- Motivation ations and plan accordingly [3, 4]. It is evident that current business data exploration and most of data visualizations are: (i) created based on task and/or data-driven models and methods; (ii) extracted based on data mining algorithms that do not consider any role-based specifications and/or user needs and requirements; and (iii) following an onesize-fits-all approach, presenting the same visualization type and content to all users irrespective of their needs, requirements, and unique characteristics.
This position paper argues that the complex nature of business processes, tasks, objectives and many data visualizations makes it indispensable to include human intelligence in the data analysis and visualization process at an early stage. It is vital to enrich the current business analytic platforms with adaptation techniques and new possibilities for interactions that will bring the human-in-theloop by considering the end-users’ individual diferences and their business context (e.g., role, purpose, requirements, tasks) in combination. It proposes a human-centred model (as the main component of an intuitive data analytics platform), that is composed of four main dimensions: User, Visualizations, Data and Tasks, and considers human factors like perceptual preferences, cognitive capabilities in information processing, afective states, domain expertise, experience, etc., amongst others. Ultimately, the goal is to enable humancentred adaptive data visualizations that will facilitate explainable exploration and transparent analysis of complex and multivariate business processes and datasets, and will support and enable more efective decision making on critical business tasks.
Today, with the growing expectations of business end-users and the proliferation of heterogeneous business processes and datasets, traditional approaches for data interpretation and visualization often cannot keep pace with the continuous escalating demand, so there is the risk of delivering unsatisfactory and misleading results. Business data models and processes characterized by significant complexity, making the analysis and understanding of data by managers, data analysts, business experts, etc., challenging, time consuming, if not many times impossible. In many cases, a single activity combines even custommade developments (e.g. using Excel) for the subsequent execution of steps creating a dispersed, inconsistent and error-prone reality.
Hence, it is widely accepted that the increasingly large amount of data requires novel, seamless, transparent and user-friendly solutions [5]. As such, handling, analyzing and gaining insights into these large multivariate processes and datasets through interactive visualizations is one of the major challenges of our days [6, 7].
In recent years, many powerful computational and statistical tools have been developed by various organizations in the business sector, such as SAS Visual Analytics1, IBM Analytics2, Microsoft Power BI3, SAP Busines Business Intelligence Platform 4, Tableau Business Intelligence and Analytics5, Qlik Business Intelligence6, etc., ofering a number of solutions like interactive maps, charts, and infographics, visual business intelligence analysis, recommend actions, etc. Interestingly [12, 13]; how individuals’ cognitive styles, like enough, these applications are currently de- Field Dependent-Independent, impact intersigned to execute the same operations follow- actions with various information visualizaing a pure machine learning approach (based tions and in relation to individual aid choices on data models and rigid tasks and objectives) and preferences [14]; or how efective are emoand with power users (e.g. data analysts) in tion-triggered (e.g., boredom and frustration) mind. They embrace the power of the sta- adaptation methods for visualization systems tistical methods to identify relevant patterns, [15]. Hence, although significant efects have typically without human intervention. Inevi- been shown in domains like public facing aptably, the danger of modeling artifacts grows plications, educational and navigation conwhen end-user comprehension and control tents, or health datasets, these ideas have rarely are not incorporated. To this end, although been applied, to our knowledge, to the busimodern business intelligence and data ana- ness sector despite the encouraging results of lytics platforms ofer vast repositories of data prior studies [16]. The current position paanalysis tools and myriads of customizable per addresses this research gap by highlightvisualizations; they have not kept up to the ing the efect of a multi-dimensional humanchallenge when it comes to their dynamic ad- centred model in data visualizations and anaaptation and personalization depending on the lytic applications that facilitate the execution role, experiences, intrinsic characteristics or of specific end-to-end business scenarios and abilities of end-users and still follow a one- tasks. The overarching innovation lies upon size-fits-all paradigm. This poses an issue as (a) the generation of knowledge and theory, the efectiveness of a visualization in terms of rules, adaptive interventions, personlization usability and understanding difers amongst conditions and explanations triggered by the users [2]. The vast amount of visual uncer- joint influence of cognitive and afective chartain information overwhelms the user’s per- acteristics on business data visualizations and ception, which in turn, severely decreases their exploration, and (b) the development of comability to understand the data and make de- putational techniques, tools and methods that cisions [3, 4]. will put the theoretical model into practice
On the other hand, the joint benefits of ad- considering the requirements, constraints and aptation and personalization, and data visu- policies of real-life business settings. alizations and exploration that consider specific human factors in the core of their user models have been highlighted repeatedly in 3. A Proposed a variety of fields and applications, mostly in Human-centred Model academia. Indicatively, research works have identified noteworthy associations of users’ This complex nature of information visualicognitive abilities like perceptual speed, in zations necessitates the development of a comrelation to performance, accuracy, and satis- prehensive theoretical model that captures imfaction when interacting with alternative da- portant factors, such as users’ cognitive charta visualization [8, 9]; others focus on opti- acteristics, afect, domain expertise and expemizing data visualizations based on the users’ rience, as well as understanding of the endgoal, behaviour, cognitive load and skills [10, user roles, objectives, context and the charac11]; investigate how human factors like per- teristics of the data [16, 17]. These factors can sonality and working memory afect user per- be utilized within the data analysis and visformance when interacting with visualizations with regard to their applicability in specific business settings and actions optimizing current multi-dimensional human-centred user models [18] that may consider factors like perceptual and cognitive processing characteristics – have an efect on the complexity of the content regarding users’ task performance, overall eficiency and cognitive control of information [19], for problem solving and comprehension during the interaction process; affect (or afective states) – referring at some extent to Emotional Arousal and Emotion Regulation, influencing people’s performance, judgeFigure 1: Proposed Human-centred Model ment and decision making process [20] while interacting with data visualizations; domain expertise and experience – directly related to ualization process to enable powerful adap- graph comprehension, accuracy and perfortation techniques generating more efective mance of users when interacting with graph interactions. In this respect, the following tasks as well as to user preference [11], sattheoretical model is proposed (see Figure 1), isfaction and the capability of being familcomprised from four main dimensions: User, iarized or switching between graphs to obVisualizations, Data and Tasks, which will be tain information; and business role – a person used as the main driver for further develop- or an entity that is defined by specific objecment and realization. tives, responsibilities and tasks and is the one that makes decisions and triggers a process, 3.1. User or specific activities, using one or more business scenarios of an organization. Data visuThis dimension is the central point of our en- alizations should be adjusted to the requiredeavour, referring on one hand to the under- ments of each role aligned to the variability standing of the business users’ roles, nature of tasks, level of knowledge, constraints, etc., and their contexts of functioning and interac- conveying the adequate information, when tion, and on the other hand to the definition and how it is needed, and on the expected of the human cognitive and afective states breadth and depth that could support and faand their transitions during the interaction cilitate a fast and accurate decision making; process with data exploration and visualiza- along with the more static ones (e.g. name, tions. As such, various interventions and adap- age, education, etc.) tive conditions may be proposed restructuring the respective contents and functionality to the needs and abilities of users, e.g., pre- 3.2. Visualizations senting more explanations, additional navi- Data visualizations are most often used to congation support and clarity, reducing the num- vey some meaning out of data and to comber of simultaneously presented stimuli and municate information. Currently, there are the volume of content. More specifically, build- diferent types of visualizations (e.g. graphs, ing upon previous research (see section 2), a plots, tables, etc.) which are used interchangenumber of human factors will be investigated ably depending on the scope and the needs of a task. For example, the typical bar and col- outcomes/ messages. At the same time, main umn charts are some of the most used visual- concern is to co-op with the risk of uncerization techniques for comparing data across tainty and data quality derived from situacategories (single or multiple), since in a co- tions where not only the types of data or feaordinate system the occurrence of a value is tures can be diferent, but there is also a varicompared directly to its neighbours; or the ety of uncontrolled efects (e.g. dependency line charts show a connection of data points to data acquisition organizations typically rein a coordinate system generating a sequence side), that could hinder the more competent of values which is used to view trends and discovery of patterns and useful information cycles over a period of time. Visualizations that in turn could enable a more efective dethat have some common and comparable fea- cision making. The mechanisms that will be tures, a recognizable impact of individual dif- considered at this stage will provide high qualferences on them, and apply at a large ex- ity business knowledge that will also detertend in the business domain will be qualified. mine (based on their properties) the signifiOnce data visualizations are defined, they and cance of the data objects and the yielded adaptheir sub-optimal counterparts will constitute tive data visualizations. This dimension will a number of subsequent objects which will be make sure that data integration is possible, enriched with metadata (semantic augmen- by means of intelligent pre-processing and tation) enabling the filtering process accord- fusion of data; to render data from diferent ing to the human-centred model and the data locations or in diferent types so to be comattributes and structure. Thereupon, adapta- parable, and to create mappings among feation and personalization techniques will be tures so that integrated data analysis will be crafted to ofer: (a) Dynamic alteration of the possible. Several unaddressed issues will be content presentation and hierarchical struc- tackled in supporting data analysis of busiture of data visualization attributes (e.g., re- ness datasets especially through the use of ordering, salience, size, saturation, texture, visualization, such as (a) very large, i.e., scalcolor, orientation, shape, etc.); (b) provision ability, (b) dynamic, i.e., addressing the veof various navigation tools and support (e.g., locity aspect within the V’s of Big Data, and visual prompts, explanations) for data/ visual (c) heterogeneous, i.e., consisting of diferexploration during end-to-end business tasks ent data types both in terms of acquisition execution; (c) variable amount of user con- method and representation [21]. trol (e.g., allowing further (deeper) data exploration); and (d) additional assistive tools 3.4. Tasks (e.g., data properties and details), etc.
3.3. Data as a sequence of actions, undertaken by the end-users. They are usually part of a wider A big challenge currently for the research com- constellation of business activities and promunity, is to develop intelligent data mining cesses that are executed with the purpose of mechanisms that can support the eficient ex- accomplishing a specific business goal (e.g., traction and fusion of multivariate data from define/ maintain material and external serdiferent locations, their integration into a uni- vices demand, or oversee stock, material deifed information model so that it can seam- mand and supply). Transparency, explainabillessly support exploratory data analysis for ity and support of end-to-end tasks execution understanding, interpreting and modelling the is a big challenge currently in the business domain, as users many times strive to un- tions and explanations? How to design and derstand the flow, dependencies and contents develop transparent and personalized condiof multi-variate information (usually gener- tions that can ensure seamless end-to-end data/ ated by diferent business processes and data visual exploration support? What kind of commodels) while at the same time are not in- putational intelligence algorithms need to be cluded in the subsequent decisions that lead developed to ensure data integration and futo a result or to actionable knowledge. Hence, sion of various dispersed datasets/ sources? recognizing the essential role of the end-user How to verify the validity of the theoretical in the data visualization and exploration pro- human-centred user model? Subsequently, a cess, it is important to enable efective human rigorous methodological approach need to be control during the tasks execution by extend- embraced, following an incremental design ing the usability and usefulness of compu- and development iterative process. The protational process models and visual analytic posed theoretical model will guide the implemethods to gain insights and value out of the mentation of an intuitive data analytics platdata towards informed business decision mak- form that will dynamically adjust (i.e., ofering. In this respect, machine learning tech- ing alternative representations) to the unique niques may be employed for analyzing big end-users characteristics, data structure and datasets arising from complex business pro- semantics of data visualizations, and explocesses and scenarios, for e.g., discovering pat- ration tactics derived from the various busiterns in data simulations or for modelling un- ness data sources/ processes of an organizacertainties increasing transparency of tasks tion. execution (e.g. change of a product’s demand).
Additionally, many problems in the business area can be formulated as probabilistic infer- 4. Expected Benefits and ence problems. Thus, a focus on probabilistic- Impact based data mining methods, including graphbased data mining, topological data mining Given the users’ diversified individual diferand other information-theoretical-based ap- ences in cognitive processing, afect, percepproaches (e.g., entropy-based), as well as on tual preferences, role, requirements, needs, the human-in-the-loop concept for increas- and expertise, as well as the size, diversity ing explainability while users interact with and processing overhead of big business data respective business use cases will be consid- sets, it is expected that this research will yield ered. lfexible best-fit data visualizations and explo
It is apparent that the successful realiza- ration methods that will support the unique tion of the above human-centred model ad- end-users with the expected transparency and heres to a number fundamental research ques- explainability during an end-to-end interactions that need to be addressed, such as: Which tion. The suggested adaptive interventions parameters and human factors are considered build on the premise that graphics and text significant so to define an inclusive human- have a complementary role in information precentred user model in the context of business sentation – while graphics can convey large data visualizations? What, how and when amounts of data compactly and support disdata visualizations content can be enriched/ covery of trends and relationships, text is much altered and delivered to the end-users? What more efective at pointing out and explaining adaptation techniques and interventions are key points about the data, in particular by fofeasible for generating best-fit data visualizacusing on specific temporal, causal and eval- exploration of data sets and processes. The uative aspects. Crafting diferent modalities aftermath is to increase the users’ understandnot only makes the presentation more engag- ing through explainable visual information as ing, but could also better suit users with dif- well as their ability to quickly act upon it while ferent cognitive abilities and afective states. engaging into purposeful transparent exploIn a broader perspective, the results of this re- rations of end-to-end business tasks. search work will have a wider business and economic impact by helping users to comprehend and familiarize themselves with usable 6. Acknowledgements data visualizations adjusted to their knowledge and abilities, enhancing their satisfaction and acceptability of related end-to-end business workflows and services. Main vision is that such practices, which provide human-centered data visualizations and visual analytic services, will be incorporated in future tools and systems, increasing the support and efectiveness of decision making in References critical tasks, enabling fast and inclusive action plans, and cutting down unnecessary iterations and costs.
Research and Innovation Foundation under the projects IDEALVis (EXCELLENCE/0918/0366) and RABIT (START-UPS/0618/0053) and the European Union under the project SLICESDS (No.951850). 5. Conclusion [1] IBM, Bringing big data to the enterprise,
accessed on jan. 1, 2020, 2016. [2] S. Liu, W. Cui, Y. Wu, M. Liu, A survey on information visualization: recent advances and challenges, The Visual Computer 30 (2014) 1373–1393. omy is to use business intelligence platforms and data analytics that ofer advanced data visualizations and exploration capabilities, beyond the traditional data-driven filtering (drilldown) and customization. The objective is to facilitate end-users (e.g., data analysts, business experts) during tasks executions to obtain insights and unveil complex patterns that lie within data and processes for making informed decisions. Accordingly, this position paper proposes a multi-dimensional humancentred model – composed of factors, such as users’ roles, cognitive individual diferences in information processing, afective states, perceptual preferences, domain expertise and experience – as the theoretical cornerstone of an intuitive data analytics platform, that may ofer seamless adaptive data visualizations and Human-Centered Issues and Perspec- forming adaptive information visualizatives, volume 4950, Springer, 2008. tion support, in: Proceedings of the [7] A. Kerren, F. Schreiber, Toward the role SIGCHI Conference on Human Factors of interaction in visual analytics, in: in Computing Systems, 2014, pp. 1835– Proceedings of the 2012 Winter Simula- 1844. tion Conference (WSC), IEEE, 2012, pp. [14] B. Steichen, B. Fu, Towards adap1–13. tive information visualization-a study [8] S. Lallé, C. Conati, G. Carenini, Impact of information visualization aids and of individual diferences on user expe- the role of user cognitive style, Fronrience with a visualization interface for tiers in Artificial Intelligence 2 (2019) public engagement, in: adjunct publi- 22. cation of the 25th conference on user [15] D. Cernea, A. Ebert, A. Kerren, A study modeling, Adaptation and Personaliza- of emotion-triggered adaptation methtion, 2017, pp. 247–252. ods for interactive visualization., in: [9] D. Toker, C. Conati, G. Carenini, UMAP Workshops, 2013.
M. Haraty, Towards adaptive infor- [16] T. Poetzsch, P. Germanakos, mation visualization: on the influence L. Huestegge, Toward a taxonomy of user characteristics, in: Interna- for adaptive data visualization in tional conference on user modeling, ad- analytics applications., Frontiers Artif. aptation, and personalization, Springer, Intell. 3 (2020) 9.
2012, pp. 274–285. [17] B. Mutlu, M. Gashi, V. Sabol, Towards a [10] B. Steichen, G. Carenini, C. Conati, task-based guidance in exploratory viUser-adaptive information visualiza- sual analytics, in: Proceedings of the tion: using eye gaze data to infer vis- 54th Hawaii International Conference ualization tasks and user cognitive abil- on System Sciences, ????, p. 1466. ities, in: Proceedings of the 2013 inter- [18] P. Germanakos, M. Belk, et al., Humannational conference on Intelligent user Centred Web Adaptation and Personalinterfaces, 2013, pp. 317–328. ization, Springer, 2016. [11] D. Toker, S. Lallé, C. Conati, Pupillom- [19] S. Lallé, D. Toker, C. Conati, G. Carenini, etry and head distance to the screen to Prediction of users’ learning curves for predict skill acquisition during informa- adaptation while using an information tion visualization tasks, in: Proceedings visualization, in: Proceedings of the of the 22nd International Conference 20th International Conference on Intelon Intelligent User Interfaces, 2017, pp. ligent User Interfaces, 2015, pp. 357– 221–231. 368. [12] T. M. Green, B. Fisher, Towards the per- [20] Z. Lekkas, N. Tsianos, P. Germanakos, sonal equation of interaction: The im- C. Mourlas, G. Samaras, The role of pact of personality factors on visual an- afect in personalized learning, in: alytics interface interaction, in: 2010 2009 Ninth IEEE International ConferIEEE Symposium on Visual Analytics ence on Advanced Learning TechnoloScience and Technology, IEEE, 2010, pp. gies, IEEE, 2009, pp. 629–633. 203–210. [21] H. Chen, R. H. Chiang, V. C. Storey, [13] G. Carenini, C. Conati, E. Hoque, B. Ste- Business intelligence and analytics: ichen, D. Toker, J. Enns, Highlighting From big data to big impact, MIS quarinterventions and user diferences: in- terly (2012) 1165–1188.