1.2 Research Problem and Approach
2.1.5 Data–Frame Model
Klein et al. [96] propose a sensemaking model that centers around data and frame.
Data is the information that a person receives or searches for, and frame is the mental structure that organizes and explains the relationship of such data. For instance, a frame can be astory, explaining the chronology of events and the causal relationships between them; or amap, showing where the events take place and the routes between them. Sensemaking is considered as a deliberate effort to understand an event, starting when a person realizes a gap of their current understanding of that event. Klein and his associates describe seven activities involved in sensemaking and are summarized in Figure 2.4.
2.1 Sensemaking 17
completion.
Empirical findings
the research it’s based on, with common beliefs. For that reason,
Data/Frame model cannot be considered a depiction of commonsense views.
Causal reasoning
Studies of domain practitioners’ stories
about how the y understoo d real-life deci- sion- making situatio ns suggest that transi- rts tion sequences—beliefs about what conve
one situation into another—are typically based on about three to four causal
factors. For example,in
explaini ng why one sports
team beat another, ne
wspaper accounts typ -
ically focus on a single ev ent such
as a cr it -
ical turnov
er (“and that cost them
the game”),
or perhaps that plus one or two other ev ents, such as a star player doing poorly or well. Giv
en the game’s length, we can see these as ov ersimplif ications, but most people would skim o ver an y account that tried to
capture a game’s full comple xity. That’s why we introduc ed the billiards metaph or earlier, to illustrat e a preferen ce for chains A sin- 2,6 al- of simple cause-effect relationships. gle
causal factor at each junction might be the preferred form of explanation, though such explanations open the
decision maker up to the reductive tendency.7
Consideration of hypotheses Decision makers are sometime s advised
that they can reduce the lik tion error by av elihood of a fixa-oiding early consideration of a hypothesis.8 But the Data/ Frame The- ory regard s early consid eration to a hypoth e-
sis as advantageous and ine
consideration vitable. Early—the rapid
recognition of a frame—pe rmits more ef ficient informati on
gathering and more specific e xpectancies that can be violated by anomalies, permit-
ting adjustment and reframing. Jenny Ru- d o l p h 9
found that decision makers must be s u f f iciently com
mitted to a frame in order to be able to test it
effecti
vely and learn from its inadequacies—something that’s
missing from open-minded and open- ended diag-nos
tic va
gabonding.Winston Sieck and his colleagu
es hav
e found that domain experts are more likely to question data
than no v- i c e s , perhaps because they’ re more famil- 10It might iar with instances of faulty data.
also mean that experts are more
conf in their frames and therefore more ident
skepti- in contrast to
cal about contrary
evidence, novices who are
less confthey ident in the frames identify
. These observa
tions would suggest that efforts to train decision mak ers to keep an open mind 11can be counterproductive, a n d e f
forts to make machines that do the va
gabonding for the human might be simi-
larly unhelpful. W
e hypothesize that meth- ods designed
to prev
ent premature consid- eration to a frame will degrade
performance under conditions where active attention management is needed (using frames) and
where people hav
e difficulty finding useful frames. Spoon-feeding interpretations to the human (via such methods as data fusion)can be
counterproductiv e.
Feedback
and lear ning
Another implication of the Data/Frame Theory concerns using feedback to pro-
mote learning. Frames are by nature reduc-
tive. And yet, frames can help overcom e the reductive tendency. The commitment
to a
frame must be coupled with a motive to tes t the frame to discover when it’s
inaccurate.
This process hinges on feedback of a certain kind. Outcome feedback (“you got it
wrong”) isn’t nearly as useful as process feedback 12because knowing t h a t
(“you did it wrong”),
performance was inadequate
isn’ t as valuab
le as understanding whatto modify in the
rea-
soning process. This includes the frame itself, because that will determine the way feedback
is understood. In other words, people need sensemaking to understand the feedback that
might improve sensemaking—the cycle as shown in figure 1. The implication is
that
people might benefit more from
intelligent systems that guide the improvement of frames e than from systems that generate alternativ understandings and hypotheses and foist them on the human.
Sensemaking as a skillWe haven’t seen evidence for a
general e
sensemaking skill. Some incidents we’v collected do suggest differences in motiv a
- tion—an “adaptive mind-set” of active
l y looking to make sense of events, as
illustra ted in essay 1’s example of the patient
with a
pacemaker. It might be possible to dev
e l o p intelligent systems that acknowledge the Pleasure Principle of human-centered com-
puting13and promote a positive motivat io n to SEPTEMBER/OCTOBER 2006www.computer .org/intelligent 89 Elaborate a frame Question a frame Track anomalies Detect inconsistencies Judge plausibility Gauge data quality Add and fill slots
Seek and infer data Discover new data/ new relationships Discard data Data Frame Reframing cycle Elaboration cycle Preserve Recognize/ construct a frame Manage attention and define, connect, and filter
the data Reframe Compare frames Seek a new frame
Figure 1. The Data/Frame Theory of sensemaking.
Figure 2.4: The data–frame model of sensemaking. It describes a set of interconnected
sensemaking activities centering around data and frame – the explanatory structure of data.
Image source: [96].
• Connect data and a frame. A person recognizes relevant pieces of data and constructs an initial frame to explain them. The frame then helps the person filter and search for new data.
• Elaborate the frame. As more is learned about the situation, the frame becomes more elaborate with new data and new relationships.
• Question the frame. The question happens when a person encounters data that is inconsistent with the existing frame. At this point, the person may be unsure that the frame is incorrect, or the data is inaccurate.
• Preserve the frame. A person may consider the severity of the inconsistent data, justify why it mismatches the frame, and ignore it.
• Compare multiple frames. Depending on experience, a person may think of alternative frames explaining the same set of data. These frames need to be compared to select the most likely one.
2.1 Sensemaking 18
• Reframe. When encountering inconsistent and contrary data, a person may need to find a replacement to explain all data. Considering discarded data and/or reinterpreting data could facilitate this activity.
• Seek a new frame. A person may deliberately search for a new frame when encountering plenty of conflicted data. One or two key data elements may serve asanchorsto help the person to elicit another frame.
The Pirolli and Card’s model describes a step-by-step process of sensemaking, in which the analyst collects relevant data and eventually transposes it into reasoning answers. However, the various sensemaking activities in the Data–Frame model may explain the strategies used by the analyst more comprehensively.