Human Factors of XR: Using Human Factors to Design XR Systems
Real World Cognition Loop for IoT
1. A Real World Knowledge Loop for
the IoT
Darminder Singh Ghataoura, Unis
2. Verifying/Formulating A Hypothesis – As part of RWK Cognition Loop “in essence”
Consists of Three Parts:
• Part 1- A Hypothesis is an explanation of an already known real world phenomena (A
posteriori – Dependent on Experience)
This can be formulated and fed directly into iCore by a Domain Expert and persisted in a RWK
store. The iCore system then has the ability to query this.
• Part 2 - A Hypothesis can also be verified independent of experience (A-priori – Verified
through reasoning alone):
This can be justified/tested by a domain expert, based on observed historical data
Hypothesis testing is a systematic way to test claims or ideas about a group or population.
Test some hypothesis by determining the likelihood that a sample statistic could have been
selected, if the hypothesis regarding the population parameter were true.
• Part 3 – A Hypothesis can be learnt/automatically derived/formulated from observed
historical data, with time, by the iCore system. (e.g. New Insights)
Might require Domain Expert Intervention
Could be used as a pre-cursor for eventual Hypothesis Testing (Important to Test Hypothesis on a data
set that was not used to generate the hypothesis, eventually data can be found to support any hypothesis
!!!)
3. iCore Domain Expert or user
(service request) wanting to test a
relationship (i.e. Based on Sarah’s
new activity regime this week
does it mean she will stay home
(A) or go to work (B)
Execution/Design Time
Training Data
Historical
Database
New Observed Data
Real World
Top –Down Approach
(Part 2)
Supervised RWK Cognitive loop: Problem Solving
Machine Learning/
Statistical Analysis
Hypothesis Store/Library
Unsupervised RWK Cognitive loop: Exploration
Design Time/Slow Evolution
Results of statistical significance
Service instance may
Query (Hyp) to assist in
formulating a service
execution or validate new
service instances and
invoke them (if allowed
SLA)
Query/Search
Service Instance (Hyp)
Verifying/Formulating A Hypothesis – As part of RWK Cognition
Loop “in essence” (Diagram exploring Parts 2 and 3)
Hypothesis Learning/Mining
Bottom-Up Approach (Part 3)
Actuation (Execution Environment - Real Time)
•
In a Top-Down Approach, a database recording past behaviour is used to verify or disapprove (i.e. test) pre-conceived
ideas, notions and hunches concerning relationships in the data observed. The Domain Expert is part of the cognitive loop and
can therefore direct and offer a better form of control on the cognitive course (i.e. informed decision making).
•
In a Bottom-Up Approach, no prior assumptions are made. Comes in two flavours. Either as Directed knowledge discovery which
attempts to explain/categorise a particular known data field or Undirected which attempts to find similarities/patterns among
different categories to infer/generate a new relationship/RW Phenomena.
4. iCore Domain Expert or user
(service request) wanting to test a
relationship (i.e.
Based on
Participant A browsing and
Participant B browsing do they like
…X
Execution/Design Time
Top –Down Approach
Training Data
Machine Learning/
Statistical Analysis
Current Real World Knowledge Model of
Query/Search
Interest (e.g. smart meeting)
Unsupervised RWK Cognitive loop: Exploration
Hypothesis Learning/Mining
(Discovery of New relationships as
Real World Facts)
Design Time/Slow Evolution
Results of statistical significance
(Execution Environment - Real Time)
Bottom-Up Approach
iCore Domain Expert or user
(service request) wanting to add
new Real World Facts to preexisting Real world Model and
Therefore “growing” RWK model
(Offer RWK model as a Service
through iCore)
Running Service Instance
Hypothesis Under Test
Add: Validated
Hypothesis
Historical
Database
New Observed Data
Real World
Supervised RWK Cognitive loop: Problem Solving
5. End User (Data Consumer/Domain Expert)
Hypothesis Validation
Hypothesis Under Test
Add: Validated
Hypothesis
Historical
Database
New Observed Real World Data
IoT World
Top –Down Approach
Machine Learning/
Statistical Analysis
Knowledge Discovery
Current Real World Knowledge Model of
Interest
Hypothesis Discovery
Discovering interesting relationships
hidden in data sets generated by the
knowledge discovery phase
Bottom-Up Approach
End User (Data Consumer/Domain Expert)