Transcript Breakout 1-personal models - Building New Theories of Human

New Computationally-Enabled Theoretical Models to Support Health BC&M
What about personal models versus general
models? How can you imagine developing and
evaluating a model that could be used in a broad
population and then adapting that model
dynamically to individuals.
Ross Hammond
James Lester
Nillo Saranummi
Brigitte Piniewski
Vicente Traver
Scribe: Jimi Huh
Breakout group summary
1
New Computationally-Enabled Theoretical Models to Support Health BC&M
Problem area #1
• Challenge/barrier: Current scarcity of individuallevel, context-aware data
– Makes construction and testing of individual-level
models very difficult
• Bold step: Create a community data commons
– linking citizens with their co-occurrence data
– Not HSS population-level data, which is not
translatable to individuals
– NOAA-like model in health, complete with IP-free
zone and supporting multiple, diverse business models
Breakout group summary
2
New Computationally-Enabled Theoretical Models to Support Health BC&M
Problem area #2
• Challenge/barrier: Continue to build models
focused on health outcomes per se rather
understanding principles of mass participation
that can produce health as by-product
• Bold step: Think broader, e.g., free parking,
micro-loans in developing countries lent to
groups of people (social pressures and contract
mechanism leveraging social dynamics)
Breakout group summary
3
New Computationally-Enabled Theoretical Models to Support Health BC&M
Problem area #3
• Challenge/barrier: Absence of pre-clinical data
and starting after problem arises, e.g., obesity
• Bold step: Collaboration to collect large volume
of co-occurrence data on an on-going, e.g.,
students providing data, whether or not they
suffer from a particular condition
Breakout group summary
4
New Computationally-Enabled Theoretical Models to Support Health BC&M
Problem area #4
• Challenge/barrier: Context of phenomena
matters, but equally important, temporal
sequences matter to capture path
dependence
• Bold step: Two approaches: 1) Data streams
and data sharing to capture sequences
longitudinally (L-EMA), 2) Process mining –
inferring context-dependent knowledge, e.g.,
prior habits; 3) Consider pathway dependence
in phenomena which can impact intervention
selection
Breakout group summary
5
New Computationally-Enabled Theoretical Models to Support Health BC&M
Problem area #5
• Challenge/barrier: 1) Previously dominant
“received” constructs for research design, e.g.,
causality, bell curve, emphasis on average,
correlational analysis, 2) Data collected for
pragmatic reasons, and questions raised based
on whatever data is available
• Bold step: 1) Adopt a greater tolerance for nontraditional approaches to modeling
phenomena being studied, 2) iteration between
model formation and data collection (iterative
model refinement, e.g., MIDAS)
Breakout group summary
6