A specific effect of using visualization,
according to [10], is the influence on the mind of an observer through the
formation of a new informational reality. The explanation of this result is
based on the study of the illusion influence of visualization objectivity on
the result of its interpretation. For systems of scientific visualization, the
usage and consideration of all the features of cognitive interpretability of
visualization are primary tasks [2, 7], which solution is complicated by the
lack of information in systematic approaches to visual analytics.
Researchers [3]
gave a rating of unsolved problems of visualization. Nowadays, this list can be
refined, taking into account current tasks, which solution can involve
visualization, including scientific and cognitive ones:
1.
Quantitative measurement. The
problem of comparing and choosing visualization tools due to an absence of a
generalized visualization scheme and the corresponding system of evaluation
criteria.
2.
Usability. The problem of achieving
by means of user characteristics visualization, providing a possibility of
obtaining a reliable solution for research problem while using only visual
interpretation.
3.
Perception. The task of
increasing the information content and interpretability of visual analytics
tools as a result of directional usage of visual perception potential.
4.
Pre-awareness. The problem of users
over-informing, which reduces cognitive effectiveness of visualization and
increases resources intensity of visualization tools.
5.
Training. The increase of
complexity of tasks of visual research and visual analytics tools creates the
need of preliminary user’s training.
6.
Scalability. The task of
finding ways to visualize information that preserves high cognitive
interpretability with any changes in detail or volume of studied data.
7.
Interpreted aesthetics. An
exclusion or usage of the subjective criteria of visual aesthetics to achieve
the goal of the study.
8.
Dynamics. The task of using
perception of changing images to enhance cognitive interpretability of
visualization.
9.
Intra-system logic. Formalization
of visualization usage as a sign language system for obtaining new knowledge
using only intersystem operations.
10. Presentation of knowledge. The solution of a
complex task of efficient storage, transmission, acquisition and use of
information as a result of creating its visual presentation.
Cognitive
interpretation of visual images imposes requirements to developers of
visualization tools, which fulfillment might be difficult, because of the need
of the directed use of perception features and thinking, many of which are not
studied well [1, 4]. In this article results obtained from the study of
features of visual perception and cognitive interpretation of abstract images
are presented.
In this article
[8], there was used a meaning of a visualization structural unit, which allows
to systematize tools designed to solve problems of various types and levels of
complexity. Combination of structural units in current system of visual
analytics, leads to the construction of a visual model, which cognitive value
exceeds the effectiveness of the study of individual images. Emergence of a
visual model is provided by links of its structural
elements.
Based on
differences in ways of the final implementation and practical use of structural
elements of a visual model, the links between them are divided into two
functional groups: informative and control links (Fig. 1). Informative links
are processes of transferring the necessary data between structural units,
which create the basis for building an image of data interpreted while solving
a task. Control communications are data management processes. Management
includes making a decision about the form of the visual image necessary to
answer the question, as well as the correspondence between the question and the
formulated hypothesis.
Based on results of solving a number of
practical tasks, a balanced model of overall effectiveness of WF
visualization was introduced, which combines positive and negative factors
characterizing visualization processes, including cognitive interpretability.
Thus, the overall effectiveness is:
WF = W+(KTrg(TA))
+ W+(KAdd(TA)) — WC (TA)
— WM (TA),
where W+(KTRG(TA))
is the value of solving the problem of analysis, W+(KADD(TA))
is the value of additional knowledge gained from formulating and testing
solution hypotheses, WÑ(TA) is the value of resources spent to create a visual model, WM(TA)-
is the value of resources used to control properties of the model in the
process of obtaining a solution.
To obtain empirical data, which confirm the
developed visualization scheme, as well as to determine factors affecting its
effectiveness, the corresponding «Visual Representation Analyzer» software has
been proposed and developed (Fig. 2). The algorithm of this software is based
on the assertion about the direct relationship between effectiveness of the
process of visual research and time taken by a user to achieve the goal of
analysis. Thus, the main task for developers of data visualization is to reduce
time of a visual research. Usage of the software " Visual Representation
Analyzer " allows to obtain data of dependence of research time at each
stage of solving the problem of analysis from any factors that have a
significant impact on it.
Obtaining and the correct interpretation of
empirical data creates the conditions for construction of visualization tools,
which have a reasonable and guided cognitive effect [5]. To achieve this goal,
a technique of studying possibilities of visual representation has been
developed, it was tested on several versions of means of visual analysis of
multidimensional heterogeneous data. The technique consists in implementing a series of solutions of test problems, accompanied
by a number of controlled restrictions, and measuring time intervals of
interaction between the researcher and the visual model. User's interaction with the model, in this case,
implies any operations available to the user, with the exception of changing
the function of the visual representation. Permissible operations are the
adaptation of data visual representation to peculiarities of user's perception,
data filtering, transition to the next task, reading attribute values of
visualized data and some others.
The test solution involves creating
an image of studied data using a predefined method of visual representation
(Fig. 3). The choice of the initial method of visualization corresponds to the
type of the research task (the task of training, informing, making a decision)
and does not depend on preferences of the researcher. Thus, at the initial
moment, a user has a visual data model and knows the purpose (question) of the
problem being solved by him [9].
While making the test solution, the
researcher, who analyzes the image of data, is allowed to formulate an
unlimited number of hypotheses to answer the problem’s question; each
hypothesis is considered as a step of analysis. The formulation of the correct
hypothesis means the completion of the test solution. The cyclical nature of
the solution procedure allows to automatically record the duration of each
stage for subsequent analysis of the effectiveness of visualization.
The obtained results allowed to make a
number of conclusions about the nature of the interaction between the user and
the visual model in the process of solving the problem of analysis. First of
all, the proposed method of studying cognitive interpretability of
visualization made it possible to estimate the total time period and influence
of a number of factors on the duration of the decision. For example, in case of
a series of measurements with participation of a single user, it made it
possible to assess individual characteristics of cognitive processes. A typical
example of results of such measurements (Fig. 3) shows a significant change in
speed of a researcher’s thinking during solving research problems of any given
type.
According to results of measurements, in
the process of the user and the visual model interaction, intervals are
allocated, where time spent by the user on building the next hypothesis is
continuously reducing. In many cases, intervals with similar signs are located
at the beginning of the analysis process, however, they may occur in further
steps of the solution. In accordance with the definition of a visualization
structural unit and the definition of the analytical visual model, similar
intervals occurred in initial steps (Fig. 4) could be defined as learning steps
that correspond to a user's familiarization
with the used visual representation function. Selecting this stage of
familiarization allows entering a qualitative assessment of the visual
representation function, which determines the consistency of perception of the
particular user and properties of the selected data visualization method.
Based on these measurements, the effect of
individual properties of the researcher on the duration of the learning
interval, which is concluded with the transition to the rapid construction of
new hypotheses, is shown. Individual characteristics of perception, according
to results of measurements, include both possibilities of visual perception and
user's prior knowledge that has an influence on results of cognitive
interpretation of visualization. Targeted use of prior knowledge of a user and
properties of his perception are ways to increase effectiveness of visual
analysis.
A change of construction time of new
hypotheses, observed throughout the analysis process, is periodic. During
solving the problem of analysis, the user did not receive information from
external sources, so a change in time spent at each step of the analysis can be
explained by a change in ratio between direct and inverse processes [6]. A
slight decrease of speed of making the decision leads to an increase in time
spent on constructing and testing the next hypothesis, and can be interpreted
as a doubt in correctness of actions, caused by errors made in previous steps.
A significant reduction in speed of analysis that arises differently depending
on a user can be interpreted as a user's need for an additional pause before
building a new hypothesis. In most dimensions, a fast increase of construction
time of a new hypothesis arises after series of quick erroneous assumptions.
That means a user does not understand an information presented as a visual
image.
|
|
|
Basing on observations of solving problems of analysis by various users, there
is an assumption, which explains the arising slowdown by the need of rethinking
mistakes. The completion of the reflection phase corresponds to an increase in
the awareness of a researcher, and its duration also depends on individual
characteristics of a researcher.
Constructing the scheme of the process of
visual research using the concept of a visualization structural unit made it
possible to develop a number of concepts and processes for answering general
visualization questions mentioned above. Some of these new results are: the
algorithm for building visualization tools, which meet requirements of a
specific problem to be solved, the methodology for conducting a visual study
aimed to increase its effectiveness, as well as a general classification of
visualization tasks. The proposed method, which use the software "Visual
Representation Analyzer", allows obtaining numerical estimates of changes
in the visualization performance, which arise from the involvement of these
results in solving practical problems of data research.
Creation of a classification of
visualization tasks gave several positive results in terms of cognitive
interpretability of visualization. Firstly, there is a simplification and
formalization of part of the processes necessary for creating new visualization
tools. Secondly, the presence of classification of visualization tasks makes it
possible to accumulate and systematize an experience of a user. Increasing a
user's prior awareness provides conditions for successful cognitive
interpretation of visualization while solving new problems.
The developed classification [8] allows to
increase effectiveness of creation or selection of visual analytics tools, which
are necessary for processing heterogeneous data, because of justification of
requirements for visualization tools (about 25% time reduction for choosing
visual analytics tools). To assess effectiveness there was used data (Table 1)
obtained during receiving test measurements. It is assumed that the use of
classification allows to determine a type of visualization task and to
accurately execute an algorithm for constructing a visual research tool using
data from a 
preliminary study. The increase in effectiveness in case of
increasing volume of source data, manifested in the reduction of analysis time,
is the basis for using visual analytics tools as tools for processing empirical
data for weakly formalized tasks.
While determining factors, which have a
significant impact on cognitive interpretation of visualization, effectiveness
of an integrated approach to visualization took a special attention [11]. It is
assumed that the use of an integrated approach reduces time of usage of a
visualization structural unit as a result of changes in a user's subjective
awareness. To obtain a numerical evaluation of the effectiveness of the
integrated approach, a comparison of the total learning interval durations and
reflection for the sequence of solutions of the same type problems were
proposed. The obtained experimental data allowed to conclude that the proposed
integrated approach to visualization and interpretation of heterogeneous data
allows to increase effectiveness (more than 40% time reduction of a hypothesis
formation) of visual analytics tools as a result of sharing computational and
cognitive resources.
There were proposed and developed tools for
obtaining experimental evaluations of effectiveness of visual analysis tools.
The obtained values allowed to determine the degree of influence of factors
included in the general definition of effectiveness of visual analysis on its
value.
Comparative values of effectiveness of
stages of creating visual analysis tools using the proposed algorithm for
constructing visual data models for various formulations of analysis tasks were
obtained. The possibility of increasing effectiveness of visual analysis as a
result of use of visual analytics tools, which ensure the interaction of
specialists from various fields of specialization and prior awareness, is
shown.
This work was funded by Russian Science
Foundation, project ¹18-11-00215.
1. Averchenkov V., Budylskii D.,
Podvesovskiy A., Averchenkov A., Rytov M., Yakimov A. Hierarchical deep
learning: A promising technique for opinion monitoring and sentiment analysis
in russian-language social networks // Proceedings of the First Conference,
CIT&DS 2015: Creativity in intellectual technologies & data science.
Volgograd, Russia, September 15 – 17, 2015. – Springer International
Publishing, Switzerland, 2015, pp. 583–592.
2. Bondarev A.E., Galaktionov V. A.
Multidimensional data analysis and visualization for time-dependent CFD
problems // Program. Comput. Softw, Vol. 41, ¹ 5, 2015, ðð. 247–252.
3. Chen C. Top 10 unsolved information
visualization problems // IEEE Comput. Graph. Appl, Vol. 25, ¹ 4, 2005, ðð.12–16.
4. Chen C., Czerwinski M. P. Empirical
evaluation of information visualizations: An introduction // Int. J. Hum.
Comput. Stud, Vol. 53, ¹ 5, 2000, ðð. 631–635.
5. Cook K. Mixed-initiative visual
analytics using task-driven recommendations IEEE, 2015, ðð. 9–16.
6. Pirolli P. Information Foraging Theory:
Adaptive Interaction with Information // Inf. Foraging Theory, 2007, ðð.1–28.
7. Pontis S., Blandford A. Understanding
“influence”: An empirical test of the Data-Frame Theory of Sensemaking // J.
Assoc. Inf. Sci. Technol, Vol. 67, ¹ 4, 2016, ðð.841–858.
8. Shklyar A. V., Zakharova A. A. Visual
presentation of different types of data by dynamic sign structures // Sci.
Vis., Vol. 8, ¹ 4, 2016, ðð.28–37.
9. Shklyar A., Zakharova, A., Zavyalov, D.,
Vekhter, E. Visual detection of internal patterns in the empirical data
Springer Verlag, 2017., ðð.215–230.
10. Bart R. Ritorika obraza // Izbrannye
raboty. Semiotika. Poehtika, M., 1989, ð. 616. (in
Russian).
11. Shklyar A.V., Zakharova A.A.
Strukturnyj podhod k vizualizacii dannyh [Structural approach to data
visualization]. Izdatel'stvo MAI-Print (Moskva) = MAI-Print publishing
house (Moscow), 2017, pp. 609–611. (in Russian).
