Concrete Entities

Definition

A concrete entity is a conceptual entity (i.e., a node) in the IVAS ontology for recording an instance of a symptom, a cause, a remedy, or a side-effect that may be identified or considered in the lifecycle of a visual analytics process, system, or workflow. The class of Concrete Entity consists of three subclasses, namely Symptom, Cause, and Remedy, which provide the high-level categorization in organizing concrete entities. Within each subclass, entities are first organized by four types of visual analytics processes, and further organized according to its highlighted keywords in the alphabetical order. Viewers can also use the "Find" button of the browser to search desired keywords within this page. Each concrete entity is mapped to one abstract entity, in square brackets, at the end of the descriptive statement. In the case of a cause entity, two abstract entities are shown in the form of [A → B], which is read as A causes B.

The instances below represent only the first step for building a comprehensive collection, and should only be considered as examples at the current stage. We hope that the visual analytics community will continue enrich this collection by adding more instances that are identified in our effort for designing, developing, evaluating and improving visual analytics processes, systems, and workflows. To make your contribution that will be acknowledged on this web site, please visit the Contribute page.

Symptoms (Side-Effects)

Statistics

  • The users are not sure about if the correlation index can be trusted, wondering if all outliers have been removed before calculation. [Stat-High-PD]
  • A current machine-centric decision process (statistics) is unreliable. [Stat-High-PD]
  • Statistical analysis has produced many measures and the users do not know which ones to use. [Stat-Low-AC]
  • All decisions are hinged upon this statistical predication. When it is off the mark by 5% or more, the consequence is serious. [Stat-High-Cost]

Algorithms

  • The automated recommendation is biased towards a type of decisions. [Alg-High-PD]
  • A current machine-centric decision process (algorithm) is unreliable. [Alg-High-PD]
  • The machine learning process does not result in a deployable model. [Alg-High-PD]
  • This search engine seems very slow with three or more keywords. [Alg-High-Cost]
  • The search results have too many false positives (poor precision) or too many false negatives (poor recall). [Alg-High-PD]
  • The users are confused by the sorted list, and they can easily see that the ordering of the scales of some events are not correct. [Alg-High-PD]
  • This topic modelling tool produces over 100 different topics every day. It is not easy to follow the evolution of the topics. [Alg-Low-AC]

Visualization

  • The animated bubble chart shows a lot of data, but the users do not know what is really happening and what they should see. [Vis-Low-AC]
  • Users cannot find (or perform visual search for) important information using the visualization easily. [Vis-High-Cost | Vis-High-PD]
  • A current human-centric decision process (visualization) is too slow. [Vis-High-Cost]
  • The axes in this parallel coordinate plot are of different linear and logarithmic scales. The users often mistake one scale for another. [Vis-High-PD]
  • Users cannot perceive the visually-encoded values accurately. [Vis-High-PD]
  • The rapid arrival of steam data disrupts ongoing visual analysis. [Vis-Low-AC]
  • The current visualization demands too much of the users’ time, especially when they have to count the number of connections. [Vis-High-Cost]

Interaction

  • The choices made by the users seem to be heavily influenced by their personal experience rather than the data in front of them. [Int-High-PD]
  • The analysts’ decision process is too slow. As soon as some became skilled and faster, they left us for better-paid jobs. [Int-High-Cost]
  • A current human-centric decision process (interaction) is too slow. [Int-High-Cost]
  • The users have to set so many parameters, and most of them are controlled by dials at 1 degree precision. [Int-Low-AC]
  • Users cannot control the scales of space, time, and fineness of categorization. [Int-High-AC]

Causes

Statistics

  • Averaging across many regions may lose too much information about regional variation, causing oversimplified decisions. [Stat-High-AC → Int-High-AC]
  • Failing to take into account all variables may cause the inaccuracy of an algorithmic decision process. [Alg-High-AC → Alg-High-PD]
  • The lack of statistical aggregation may be the reason why the visual search task is so slow in this visualization process. [Stat-Low-AC → Vis-High-Cost]
  • The means used to compare these three groups are statistically insignificant, and there are some risks in this conclusion. [Stat-High-PD → Int-High-PD]
  • The difference between the statistical properties of the training data and those of the data being analyzed may cause the inaccuracy of a machine-learned algorithm. [Stat-High-PD → Alg-High-PD]

Algorithm

  • This algorithmic model computes the binary relations between books based on many factors, the edges in the graph are erroneous. [Alg-High-AC → Vis-High-PD]
  • The algorithm presents the users with too many choices, and the users have to read a lot of texts and press a lot of buttons. [Alg-Low-AC → Int-High-Cost]
  • Algorithmic search shows too many options, causing slow visualization processes. [Alg-Low-AC → Vis-High-Cost]
  • An erroneous data pre-processing process may cause the inaccuracy of an algorithm. [Alg-High-PD → Alg-High-PD]
  • The errors of this recommendation algorithm may cause the poor decisions by the users. [Alg-High-PD → Int-High-PD]
  • This machine-learned model is so expensive to update. It may cause many bad decisions in this changing environment. [Alg-High-Cost → Alg-High-PD]

Visualization

  • The cluttered translucent edges in the graph cause users to spend too much time in find a path between two given nodes. [Vis-High-PD → Vis-High-CT]
  • Some visual analysis tasks (visualization) may demand high cognitive load, causing slow decision processes. [Vis-High-Cost → Int-High-Cost]
  • Some visual representations may incur high cognitive load, causing slow visualization processes. [Vis-High-Cost → Vis-High-Cost]
  • The lack of overview visualization makes users to make biased decisons based on their partial knowledge about the data. [Vis-Low-AC → Int-High-PD]
  • The lack of suitable sports video visualization, the coaches often make decisions based on their first impression of the match. [Vis-Low-AC → Int-High-PD]
  • Too much data may cause slow visualization processes. [Vis-Low-AC → Vis-High-Cost]
  • Because visual counting is costly for these analytical tasks, the users just make rough estimations, leading to some bad decisions. [Vis-High-CT → Vis-High-PD]
  • Humans’ working memory cannot store much visual information, making the users to watch these animations repeatedly. [Vis-High-AC → Vis-High-Ct]

Interaction

  • The users seems to chose only the first two (out of eight) variables to visualize , and the visualization could be biased. [Int-High-AC → Vis-High-PD]
  • The cognitive load for memory recall is so high in this analytical process, when they are rushed, their judgements are error-prone. [Int-High-Cost → Int-High-PD]
  • Different users want different variables to be depicted in the glyph designs, we end up with an overly-complex glyph design . [Int-Low-AC → Vis-Low-AC]
  • The users select parameters by trial-and-error, and the algorithmic model produces very inconsistent classification over a period. [Int-High-PD → Alg-High-PD]
  • A poorly designed user interface may cause slow interaction processes. [Int-High-PD → Int-High-Cost]
  • Poorly tuned parameters may cause the inaccuracy of a parameterized algorithm. [Int-High-PD → Alg-High-PD]

Remedies

Statistics

  • To reduce the averse consequences due to some statistical measures by bringing assistance from other VA components. [Stat-Low-Cost]
  • To provide the users with ways to inspect daily performance in addition to weekly and monthly performance. [Stat-Low-AC]
  • To remove the seasonal effects in the time series. [Stat-Low-PD]
  • To introduce more statistical aggregation of the data. [Stat-Low-PD]
  • To compute a trend line for these data points. [Stat-High-AC]

Algorithm

  • To introduce algorithmic filtering, clustering, summarization, etc. [Alg-High-AC]
  • To introduce algorithmic support for focus-and-context visualization. [Alg-High-AC]
  • To develop a GPU version of the algorithm to speed up the analytical processes. [Alg-Low-Cost]
  • To update the machine learning model with new training data that reflect the current environmental conditions better. [Alg-Low-PD]
  • To sort these data objects before visualize them. [Alg-High-AC]
  • To accompany each classification decision with an uncertainty value. [Alg-Low-AC]

Visualization

  • To avoid using emotive colors, shapes, and glyphs in displaying these politically-sensitive data in disseminative visualization. [Vis-Low-PD]
  • To avoid using emotive colors, shapes, and glyphs in displaying these politically-sensitive data in disseminative visualization. [Vis-Low-Cost]
  • To introduce focus-and-context (highlighting focus) visualization. [Vis-High-AC]
  • To introduce focus-and-context (adding context) visualization. [Vis-Low-AC]
  • To introduce overview first and details on demand. [Vis-High-AC]
  • To replace a 1D sorted list with a 2D sorted view of visual objects sorted by 2 keys. [Vis-Low-AC]
  • To add dynamic tooltip views to glyphs compensate for the over-abstraction of glyph representations. [Vis-Low-AC]
  • To introduce transitional animation for users to follow the movement of data objects better. [Vis-Low-AC]
  • To provide a treemap view to show where search results are distributed in the search space, assisting in the refinement of the search criteria. [Vis-Low-AC]
  • To introduce more visual abstraction, such as glyph-based data objects in spatiotemporal visualization. [Vis-High-AC]
  • To introduce visualization in a decision process to complement algorithmic recommendation of decisions. [Vis-Low-AC]
  • To introduce visualization to support some processes in a machine learning workflow, e.g., by using the VIS4ML ontology to identify such processes. [Vis-Low-AC]

Interaction

  • To address users’ biases by introducing algorithmic recommendation and uncertainty visualization. [Int-Low-PD]
  • To reduce users’ cognitive load in viewing videos by introducing video visualization. [Int-Low-Cost]
  • The users are given more control over the visualization processes, including zooming, filtering, highlighting, etc. [Int-Low-AC]
  • To introduce a user interface for users to set parameters for an algorithm. [Int-High-AC]