Mining the Bibliome - Translational Informatics: Realizing the Promise of Knowledge-Driven Healthcare - page 72

5.1.1 Hypothesis Generation and Hypothesis Testing

Without context or purpose, data are essentially meaningless. To elicit the value of
data, they must fi rst be transformed into information , which in turn needs to be
coalesced into knowledge before any utility may be ascertained. This knowledge,
which may be deemed as “actionable,” may be immortalized as wisdom that is used
to guide future encounters with data. Formally, the process of data transformation is
cast as the “Data-Information-Knowledge-Wisdom” (DIKW) framework. The
DIKW framework, and its contemporary incarnations largely attributed to Ackoff
[ 1 ], provides a formal construct to analyze data transformation. Most importantly,
the DIKW framework offers both context and purpose as constraints to instill mean-
ing into volumes of data. This architecture is increasingly important as improve-
ments in data generation and acquisition technologies continue to exceed intellectual
capacity for interpretation. It is outside the scope of the present discourse to describe
the complete process of data transformation associated with the DIKW framework.
Nonetheless, the DIKW framework offers a useful construct to describe how data
are used within biomedical contexts.
The advancement of technologies across the spectrum of biomedicine has
resulted in a new cadre of data that are referred to as “Big Data” Chap. 7 provides
more detail about the nuances of Big Data. Within the current context, where the
focus is to leverage knowledge that have been recorded in some reusable form,
the discussion will be around approaches that are used for one of two purposes:
(1) for hypothesis generation ; or (2) for hypothesis testing . Historically, these
two purposes can be perceived to be in confl ict with each other; the increased
ability to generate hypotheses is of no value without completing the testing of
those hypotheses that have already been postulated. Indeed, the scientifi c meth-
odologies for hypothesis generation and testing are largely considered indepen-
dently (those that generate hypotheses seldom actually test them and those that
test hypotheses seldom are focused on hypotheses generation approaches).
However, the realities of contemporary scientifi c inquiry in light of the volume of
data that are available require a synergy between the generation and testing of
hypotheses.
For the purposes of this chapter, it is not essential to fully understand the
philosophical principles of the Baconian Method or Scientifi c Method, which
can be seen as frameworks for respectively formalizing the process of hypothesis
generation and testing [ 2 ]. Instead, it is useful to consider these as two major
scientifi c philosophies as approaches that involve the use of data. Big Data might
be best leveraged through a Baconian process of reduction of highly complex,
highly produced, and highly heterogeneous data into tractable units of “action-
able knowledge.” Similarly, actionable knowledge might be best utilized if sub-
jected to the Scientifi c Method for validation. Thus, in the context of a learning
healthcare system, there is a necessary synergistic relationship between the
Baconion Method and the Scientifi c Method that marry the realms of hypothesis
generation and testing.