Headers will be added later
The rise of informatics as a research domain
Peter Fox1
1Tetherless
World Constellation, Rensselaer Polytechnic Institute, 110 8th St., Troy, NY 12180 USA,
pfox@cs.rpi.edu
Abstract: Over the past five years, data science has emerged as a means of
conducting science over many disciplines and domains. Accompanying this
emergence is the realization that different informatics approaches, i.e. the science
of data and information underlying the developments, have also emerged
independently and empirically in many areas, e.g. astro, bio, geo, hydro, ocean
and over different timescales, funding models and corresponding appreciation by
their communities. To fully enable both interdisciplinary research and cope with
increasingly complex data within domains, a move to a more repeatable and
interworkable mode is required, i.e. adding a research component to the
application component of informatics that enables data science as a means to
address integrative science grand challenges areas such as water, environment
and climate, ultimately resulting in the discovery of new knowledge. This
contribution details some key elements of research informatics, the class of people
who appear in this discipline, and the state of some current research challenges.
Keywords: Informatics; Data Science; Provenance; Semantics.
1
EVOLVING CONDUCT OF SCIENCE
Recent advances in data acquisition techniques quickly provide massive amounts of complex
data characterized by source heterogeneity, multiple modalities, high volume, high
dimensionality, and multiple scales (temporal, spatial, and function). In turn, science and
engineering disciplines are rapidly becoming more and more data driven with goals of higher
sample throughput, better understanding/modeling of complex systems and their dynamics, and
ultimately engineering products for practical applications. However, analyzing libraries of
complex data requires managing its complexity and integrating the information and knowledge
across multiple scales over different disciplines. For example, the reductionist approach to
biological research has provided an understanding of how the linear arrangement of nucleotides
encode the linear arrangement of amino acids and how proteins interact to form functional
groups governing signal transduction and metabolic pathways, etc. But at each level of
biological organization, new forms of complexity are encountered such that we’ve taken our
biological machines apart but can’t put them back together again (B. Yener, personal
communication) - our ability to accumulate reductionist data has outstripped our ability to
understand it. Thus, we encounter a gap in the structure/function relationship; having
accumulated an extraordinary amount of detailed information about biological, material, and
environmental structures, we cannot assemble it in a way that explains the correspondingly
complex functions these structures perform. Educating and training data scientists with the
necessary knowledge and skills to close the data - knowledge gap becomes a crucial
requirement for successfully competing in science and engineering.
Attention to data science has spread from being a discussion among researchers (Baker et al.,
2008, Nativi and Fox, 2010), e.g. the Fourth Paradigm publication (Hey et al., 2009), into more
general audiences, for example the recent Nature and Science special issues on “Data”. Not
surprisingly there is explicit emphasis on data and information in professional societies and
international scientific unions (SCCID, 2011), in national and international agency programs,
foundations (for example the Keck Foundation and the Gordon and Betty Moore Foundation)
and corporations (IBM, Slumberger, GE, Microsoft, etc.). Surrounding this attention is a
Headers will be added later
proliferation of studies, reports, conferences and workshops on Data, Data Science and
workforce. Examples include: “Train a new generation of data scientists, and broaden public
understanding” from an EU Expert Group (Riding the Wave, 2010), “…the nation faces a critical
need for a competent and creative workforce in science, technology, engineering and
mathematics (STEM)...” (NSF 2011a), "We note two possible approaches to addressing the
challenge of this transformation: revolutionary (paradigmatic shifts and systemic structural
reform) and evolutionary (such as adding data mining courses to computational science
education or simply transferring textbook organized content into digital textbooks).”
(NSF2011a), and “The training programs that NSF establishes around such a data
infrastructure initiative will create a new generation of data scientists, data curators, and data
archivists that is equipped to meet the challenges and jobs of the future." (NSF 2011b).
Currently, there are very few graduate education and training programs in the world that can
prepare students in a cohesive and interdisciplinary way to be productive data scientists. The
emphasis on advanced degree programs thus required a rich research agenda, one that draws
on many disciplines.
2
2.1
INFORMATICS – BALANCING RESEARCH AND APPLICATION
Definition
One of the more useful definitions is Informatics: “Information science includes the science of
(data and) information, the practice of information processing, and the engineering of
information systems. Informatics studies the structure, behavior, and interactions of natural and
artificial systems that store, process and communicate (data and) information. It also develops
its own conceptual and theoretical foundations. Since computers, individuals and organizations
all process information, informatics has computational, cognitive and social aspects, including
study of the social impact of information technologies.” (Wikipedia, italics added by author to
note added content). Many disciplines, science, engineering, theory, all together.
2.2
Key elements of informatics
Since science informatics efforts have emerged largely in isolation across a number of
disciplines, it is only recently that certain core elements have been recognized as increasingly
common. Since informatics bridges to end-use, the ‘use case’ (or user scenario) is one of the
most important. Introduced by Jacobson (1987) and popularized by Cockburn (2000) for the
purposes of software development, a use case ‘… is a prose description of a system’s behavior
when interacting with the outside world.’ An excellent coverage of the topic is presented by
Bittner and Spence (2002) who outline key factors in the development of use cases, especially
the ‘why’ in addition to the ‘what’ and ‘how’. The second core element is the use of conceptual
information models, especially working with domain science, application and data experts and
matching and leveraging those models with others, especially ones that underlying key
standards (see Research Challenges). There are other elements but these two are critical.
Thus an overall skillset for an informaticist features two main elements: for information
architecture - to sustainably generate information models, designs and architectures – and for
technology development - to understand and support essential data and information needs of a
wide variety of producers and consumers. In order to continue to meet evolving end-use needs
at the current rate technical and technology change, informaticists must be grounded in the
underpinnings of informatics, including information systems theoretical methods and best
practices. This combination of knowledge and skills currently resides in ~ 1 in 100 information
practitioners in the sciences (purely empirically and ethnographically determined by the author).
The balance of research, application and a strong interdisciplinary educational curriculum is
needed for informatics to address integrative science grand challenge problems that are the
centre-piece of many international research organization missions, especially in Earth system
sciences.
Headers will be added later
3
EXAMPLAR APPLICATION IN A WIRADA
CONTEXT
Sensor
Agency Activity
Sensing/
Collecting
Observations
Deploy
Sensor
<<information>>
Raw Observations
from Agency
Transfer Data
<<information>>
Sensor Information
Data Transformation
& Publishing
<<information>>
Published Observations in O&M
Kepler – Generate
Gridded Rainfall Data
Figure 1: Map of Tasmania with place marks
indicating stream gauge sensor placements.
Different colours represent different operating
agencies. The South Esk area is in the ENE.
<<information>>
Flow Forecast
Output (in TSF)
<<information>>
Specified Rainfall Range
<<information>>
Gridded Rainfall Data
Stream
Forecast Model
During 2010-2011 the CSIRO Tasmanian
Information and Communication Technologies
Centre
(ICTC)
undertook
an
excellent
<<information>>
Data
application of many of the aforementioned
Published Flow Forecast
Transformation &
Result (in O&M)
Publishing
informatics capabilities and in doing so
encountered many of the research challenges
discussed below. They developed many use
cases, information models, evaluated language
Figure 2: Information model for SEFF
encodings for sensor networks, water domain
vocabularies and provenance, and deployed the beginning with the sensor and ending with a
stream flow forecast. Rounded rectangles
software implementation in support of two
represent
processes and square rectangles
hydrological sensor webs for the South Esk
represent data and information products.
(north-east Tasmania, see Fig. 1) Flow Forecast
(SEFF; see Fig. 2) and the Water Information Research and Development Flood Early Warning
System (WIRADA-FEWS). Detailed descriptions of these systems are available in other papers
both at the conference and in the proceedings.
4
RESEARCH CHALLENGES AND PATHS FORWARD
Seen collectively, prominent domains of informatics (DeRoure and Goble, 2006) have two key
things in common: i) a distinct shift towards systematic methodologies and corresponding shift
away from the tight dependence on technologies and ii) the importance of a multi-disciplinary
and collaborative approaches. Together, these changes point to a maturing of the discipline
itself, with the aim of providing reproducible results, i.e. ones based on a solid underlying set of
theoretical concepts rather than trial and error with various technologies. The current evolution
in what capabilities are developed today is that many of the information systems theories
(Shannon, 1948), semiotics (Peirce 1898), and cognitive and architectural design principles
(Dowell and Long, 1998) are all beginning to coalesce in the modern practice of informatics.
The key new element is the fourth paradigm of science noted earlier. The problems to be solved
are more about integration, crossing disciplines and responding to a much wider variety of
Headers will be added later
stakeholders (SCCID 2011); those that really need to use many forms of data and information.
In practice this means that at the intersection of prior theoretical concepts (almost all of them
developed in the analog world of data and information) and real applications is a research
environment that is providing fertile for the new breed of informaticists and in turn, data
scientists (Hey et al., 2009). Of the many current research challenges, several examples, which
are applicable to water/hydro informatics, are discussed below.
Where’s the data and what happened to it? One of the most important and increasingly
evident challenges arises from the shift of where and how the data are being made available.
More and more ‘distance’ in the form of data and information from different computers,
undergoing network transmission and format translation means that without very careful
curation of these processes, obfuscation results and the probability of information loss
increases (Weaver and Shannon, 1963). The present worst-case scenario is that data may be
used for a purpose that it is not fit for. While it is highly likely that this situation occurs frequently
today, the consequences of incorrect conclusions or erroneous decisions are far reaching,
especially when avoidable with a combination of methodological and technical means. The
challenge then is to propagate relevant knowledge, information and/or metadata (or pointers to
them) along with the data as it a) proceeds through a workflow, b) through an analysis pipeline,
c) is extracted, transformed, and loaded into applications or transmitted over the Internet. The
ultimate realization then is that an informatics task involves constructing when needed, a sense
of ‘state’ from a highly distributed and non-uniform quality, knowledge base. As a whole, this
important topic falls under the heading of provenance research, and has grown in significance
and understanding over the last ~ 10 years (Buneman et al., 2001; Simmhan et al. 2005).
Machine processable provenance also provides and opportunity to instrument the scientific data
‘enterprise’ enabling explanation and verification use cases.
Allowing for unintended use? A closely related challenge to the previous one, also focused
on legitimate end use, is how to effectively increase the return on scientific investment made in
generating and processing data by enabling use in a discipline or application area different from
the original purpose? This challenge adds a dimension beyond the view of provenance noted
above to include additional information and documentation along the lines of what is intended
for digital data preservation. One of the biggest examples has occurred with the proliferation of
‘geobrowser’ applications (e.g. Microsoft Visual Earth tm or Google Earthtm) making immense
amounts of previous inaccessible geospatial data (often on hard copy maps or in proprietary
Geographic Information Systems (GIS) available and useable for very different purposes
(Hayes et al., 2005), e.g. navigation and feature identification or recreation in the case of the
general public, all the way to post-natural disaster damage evaluation. Past approaches, which
fall into the category of ‘save everything known about the data’ are increasingly untenable and
are unlikely to scale at the present rate of increase of data volumes and diversity. While
planning for the unknown is difficult, the small number of current strategies used (e.g. using
common/ standard data formats and metadata representations) must be augmented.
New tools or new again? As informatics approaches to data science start to make their way
into the modern design and implementation of information systems, some new opportunities for
further enhancing the conduct of science emerge. The most obvious target for change is the
user’s (presumably) electronic device that they use to perform a task. Progress with relatively
light-weight clients (hand held devices, Web browsers) has meant that people can change the
tools they are using for certain tasks relatively frequently, e.g. communicating, seeking
information, reviewing it, and perhaps retrieving a sample. This situation is desirable, as the rate
of technical change means this market sees the greatest innovation and capability
enhancement. However, as science or application research proceeds, different tools are used.
Analysis tools such as electronic spread sheets (e.g. Microsoft Exceltm) or fourth generation
language tools (e.g. Matlabtm) and some of the most common tools, are the least flexible and
least integrated into newer information systems, and most importantly, able (or not) to handle
increasing data volumes, complexity and heterogeneity (let alone semantics; see below).
Continuing research is needed to understand how common modes of scientific investigation:
Headers will be added later
analysis, modelling, visualization can be brought closer to the end user and sooner in the
process, especially for visualization (Fox and Hendler, 2011).
A role for standards? Standards, both community and those certified by national and
international bodies (e.g. the International Standards Organization; ISO), are one means for
disparate information systems to reliably exchange data and information, over the Internet
(Percival, 2010; Woolf, 2010). However, while the development of standards is largely a
methodological and technical exercise (organizational and political factors aside), the adoption
and/ or implementation of standards is very much a social (and organizational/ political) one.
Further, the need to be able to consistently but rapidly evolve standard means of data and
information exchange is often hard to achieve without paying attention to the socio-technical
aspects of the systems, or systems-of-systems. Again, informatics approaches are a valuable
path forward as the pragmatics of end-use and real implementations tends to drive different
choices in deciding what needs to be agreed on and what does not (E. Christian, personal
communication).
Encoding meaning? As noted by Baker et al. (2008), the gap that informatics bridges is that
between science and application areas and underlying, discipline-agnostic ICT infrastructure
(such as databases, web servers, wikis, etc.). The informatics task then is to model and
represent the semantics expressed in the use case, or at the very least in the questions that
need answers (Fagin et al., 2005). Traditionally, this knowledge has been embedded in
software implementations that operate on data and information, often with convoluted or
complex logic and human reasoning. There are many consequences but two prominent ones for
this discussion are the inability to present sufficient provenance and explanation as to what
happened and why (see earlier discussion; Buneman et al. 2001) and the rigid nature of the
encodings which make them harder to maintain, especially as technology changes and different
people, with different understandings of the semantics become involved. The latter
circumstance highlights the important distinction between closed world and open world
semantics, especially for data exchange (Libkin and Sirangelo, 2011). In short, the closed world
assumption (CWA) means all knowledge is known and encoded and that which is not encoded
is assumed false. A primary example is a relational database schema. In contrast, the open
world assumption (OWA) fosters incomplete and evolving knowledge encodings and what is not
encoded is simply not known. As a result another shift is occurring toward more explicit
(declarative) semantics in the form of ontologies (Gruber, 1993) and consequent use of
semantic web technologies (Berners-Lee et al. 2001) with applications in diverse science areas
(Fox et al. 2009). The challenges here are numerous (Fox and Hendler, 2009) and include
forming and sustaining suitable collaborative teams that include domain and data experts,
information and knowledge modellers, and software engineers, balancing what semantic
expressivity to encode, with how they will be implemented and used, as well as how they are
maintained and/or evolve (see standards discussion).
ACKNOWLEDGMENTS
The author acknowledges the invitation and support from the conference organizers (CSIRO)
and a Research Collaboration agreement between CSIRO/TasICTC and RPI and stimulating
research discussions with Stephen Giugni, Kerry Taylor and Andrew Terhorst. The hydroinformatics application contributions to the sensor webs mentioned in this paper are due to
Quan Bai, Corne Kloppers, Brad Lee, Qing Liu, Chris Peters and Peter Taylor (and others); they
are the informaticists of today.
REFERENCES
Baker D, Barton C, Peterson W and Fox P (2008) Informatics and the 2007–2008 Electronic
Geophysical Year. Eos Trans. AGU 89(48), 485-486.
Bittner K and Spence I (2002) Use Case Modeling, Addison-Wesley Professional.
Headers will be added later
Buneman P, Khanna S and Wang-Chiew T (2001) Why and Where: A Characterization of Data
Provenance, Lecture Notes in Computer Science, 1973, 316-330, DOI: 10.1007/3-54044503-X_20
Cockburn A (2000) Writing Effective Use Cases, Addison-Wesley, Boston, MA.
De Roure D and Goble C (2009) Software Design for Empowering Scientists, IEEE
Software, 26, (1), pp. 88-95.
Dowell J and Long J (1998) A conception of the cognitive engineering design
problem, Ergonomics , 41 (2) 126 - 139.
Fagin R, Kolaitis Ph, Miller R, Popa L (2005) Data exchange: semantics and query answering.
Theor. Comput. Sci. 336(1): 89–124.
Fox P, McGuinness DL, Cinquini L, West P, Garcia J, Benedict JL and Middleton D (2009)
Ontology-supported scientific data frameworks: The Virtual Solar-Terrestrial Observatory
experience, Computers & Geosciences, 35 (4), 724-73.
Fox P and Hendler J (2009) Semantic eScience: Encoding Meaning in Next-Generation Digitally
Enhanced Science, in The Fourth Paradigm: Data Intensive Scientific Discovery, Eds. T.
Hey, S. Tansley and K. Tolle, Microsoft External Research, pp. 145-150.
Fox P and Hendler (2011) Changing the Equation on Scientific Data Visualization, Science, 331
(6018)
pp.
705-708,
DOI:
10.1126/science.1197654
online
at
http://www.sciencemag.org/content/331/6018/705.full
Gruber TR (1993) Toward Principles for the Design of Ontologies Used for Knowledge Sharing,
Formal Ontology in Conceptual Analysis and Knowledge Representation, eds. N. Guarino
and R. Poli, Kluwer Academic Publishers, pp. 1-23.
Hayes, GS, Mayers SA and Pensack CE (2005) Using GIS to Find New Uses for Old Data,
Government engineering, Sept.-Oct. 2005, pp. 16-19.
Hey T, Tansley S and Tolle K, Eds. (2009) The Fourth Paradigm: Data Intensive Scientific
Discovery, Microsoft External Research.
Jacobson I (1987) Object oriented development in an industrial environment, OOPSLA ‘87:
Object-Oriented Programming Systems, Languages and Applications. ACM SIGPLAN, pp.
183-191.
Libkin L and Sirangelo C (2011) Data Exchange and Schema Mappings in Open and Closed
Worlds, Journal of Computer and System Sciences, 77 (3), 542-571.
Nativi S and Fox P (2010) Advocating for the Use of Informatics in the Earth and Space
Sciences, Eos Trans. AGU 91(8), 75-76, doi: 10.1029/2010EO080004.
Peirce CS (1898) Reasoning and the Logic of Things: The Cambridge Conference Series of
1898, Ed. K Ketner, Harvard University Press (January 1, 1992).
Percival G (2010) The Application of Open Standards to Enhance the Interoperability of
Geoscience Information, International Journal of Digital Earth, 3, Suppl. 1, 14-30, doi:
10.1080/17538941003792751
Report Of The National Science Foundation Advisory Committee For Cyberinfrastructure Task
Force
On
Cyberlearning
And
Workforce
Development
(2011a)
http://www.nsf.gov/od/oci/taskforces/
Report Of The National Science Foundation Advisory Committee For Cyberinfrastructure Task
Force On Data and Visualization (2011b).
Riding the wave (2010) How Europe can gain from the rising tide of scientific data http://www.grdi2020.eu/Pages/Unlock.aspx, Final report of the High Level Expert Group on
Scientific Data A submission to the European Commission October 2010.
SCCID Interim Report (2011) The International Council for Science’s Strategic Coordinating
Committee on Information and Data Interim Report, ICSU, April 2011. On the web at
http://www.icsu.org/what-we-do/committees/information-data-sccid/?icsudocid=interim-report
Shannon, CE (1948) A Mathematical Theory of Communication, Bell System Technical Journal,
27, pp. 379–423 & 623–656, July & October, 1948.
Simmhan YL, Plale B and Gannon D (2005) A survey of data provenance in e-science, ACM
SIGMOD, 34 (3), 31-36, doi: 10.1145/1084805.1084812
Weaver W and Shannon CE (1963) The Mathematical Theory of Communication. Univ. of
Illinois Press. ISBN 0252725484.
Woolf A (2010) Powered by Standards – New Data Tools for the Climate Sciences, International
Journal of Digital Earth, 3, Suppl. 1, 85-102, doi: 10.1080/17538941003672268.