Protagonists: Henry Ford, Nikola Tesla $...

Post on 10-Jun-2020

0 views 0 download

transcript

11-­‐04-­‐18  

1  

Cognitive Task Analysis Carolina Wannheden, Doctoral student Health Informatics Center, LIME

carolina.wannheden@ki.se

”The expert’s bill”

Protagonists: Henry Ford, Nikola Tesla

$ 10,000

Marking wall: $ 1 Knowing where to mark: $ 9,999

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

2  

Discuss in pairs…

1. What is it that Tesla knows, and how does he know it?

2. How can we grasp this knowledge?

3. Why would we want to grasp this knowledge?

2011-04-18 carolina.wannheden@ki.se

Learning Outcomes

  You should be able to

 Explain when and why to perform a Cognitive Task Analysis  Explain what distinguishes an expert and when knowledge-based

systems may be useful  Explain different knowledge structures  Apply the Critical Decision Method for the aquisition of expert

knowledge to inform the design of a CDSS

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

3  

The purpose of CTA…

…is to understand how people think and what they know (cognitive) to achieve some particular goal (task)…

…in order to   Analyze incidents   Develop training material   Develop memory aids, decision aids, expert systems

2011-04-18 carolina.wannheden@ki.se

When to use knowledge-based systems

  When real experts are

 Scarce  Expensive  Inconsistent  Unavailable on a routine basis

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

4  

Case from week 1

2011-04-18 carolina.wannheden@ki.se

Read the case and reflect on…

1.  What makes the anesthesiologist’s decision challenging?

2.  Why does the anesthesiologist choose the general anesthesia?

3.  How would you describe the context? Which factors affect her decision-making?

4.  What type of knowledge does the anesthesiologist need in order to make the right decision? Can you distinguish different types of knowledge?

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

5  

Experts know how, not just what!

  Conceptual/factual knowledge (what)  General anesthesia is for…  Regional anesthesia is for…

  Procedural knowledge (how)  IF X, then provide the appropriate anesthesia and operate.

  (Structural knowledge  Guideline A is applicable for patients aged 15 or more)

2011-04-18 carolina.wannheden@ki.se

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

6  

Key aspects of CTA

1. Knowledge acquisition (KA) 2. Data analysis 3. Knowledge representation

2011-04-18 carolina.wannheden@ki.se

What CTA tries to capture

  What people are thinking about   What they are paying attention to   The strategies they are using in making decisions   What they are trying to accomplish   What information they discard   What they know about the way a process works

(Crandall, Klein, Hoffman, 2006)

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

7  

Techniques for knowledge acquisition

  Interviews  + Require a minimum level of resources  + Can be performed in a relatively short time frame  + Can yield a significant amount of qualitative knowledge  - Lack of quantitative data  - Bias due to selection of questions by researcher  - Elicited knowledge may not correspond to what expert actually

does   Think-aloud protocols   Observations

 Ethnographic evaluations to collect information in context   Group techniques (e.g. brainstorming)

2011-04-18 carolina.wannheden@ki.se

Data analysis methods

  Protocol and discourse analysis  Elicit knowledge from individuals while they are engaged in

problem-solving or reasoning tasks  Determine conceptual entities and relationships between them

  Concept mapping  Node-link structures of knowledge  Concept maps can support the formation of consensus among

experts   Verification and validation

 Verification: fulfillment of perceived requirements (to define design)  Validation: fulfillment of realized requirements (upon

implementation)

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

8  

Knowledge Representation

  Narrative formats   Chronologies   Data organizers   Process diagrams   Concept maps

(Crandall, Klein, Hoffman, 2006)

2011-04-18 carolina.wannheden@ki.se

Decision Requirements Table

Treatment phase

Decision challenge

Cue/ Information

Strategy or practice

Novice Traps

Giving anesthesia before surgery

Choose adequate type of anesthesia

Age, medical history

2011-04-18 carolina.wannheden@ki.se

(Crandall, Klein, Hoffman, 2006)

11-­‐04-­‐18  

9  

Challenges in acquiring expert knowledge

  Complex and resource-intensive   Identification and access to domain expert with

 Sufficient domain knowledge  Interest in participating in knowledge acquisition process  Minimal bias

(Greenes, 2006)

2011-04-18 carolina.wannheden@ki.se

Reasoning biases

  Poor estimation of probabilities (Probability bias)  Use terms like ”suggests”, ”supports”, ”goes against”, ”often”,

”evokes the possibility” to describes uncertainty

  Estimation bias  Recency bias (mistaking for frequeny)  Anchor judgments on initial estimates  Familiarity or stereotypic frequency over objective frequency  Overestimate frequency of rare events

(Greenes, 2006)

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

10  

Critical Decision Method (CDM)

  CDM was created to learn from specific incidents (Hoffman et al., 1998)

  Described well in chapter 5, Working Minds: A Practitioner's Guide to Cognitive Task Analysis (Crandall, Klein, Hoffman, 2006)

2011-04-18 carolina.wannheden@ki.se

The CDM interview

  Intensive in-depth interview (duration ~2 hrs) to elicit cognitive functions such as decision making, planning, sensemaking within a specific challenging incident

  Conducted by 2 researchers 1.  Primary facilitator (and note-taker) 2.  Note-taker and time-keeper

  Conducted in 4 sweeps (phases) 1.  Incident identification 2.  Constructing a timeline 3.  Deepening 4.  ”What if” queries

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

11  

Sweep 1: Incident identification

  Goal: Try to identify an incident that will contain cognitive components beyond background and routine procedural knowledge  Nonroutine, challenging events  The participant has to have a role as a ”doer/decision maker”  The participant’s decision making should have had a direct impact

on the outcome  Critical event, time pressure

  Ask the participant to provide a brief account of the story, from beginning to end

2011-04-18 carolina.wannheden@ki.se

Sweep 2: Constructing a timeline

  Goal: Get a clear, refined, and verified overview of the incident structure, identifying key events and segments.

  The interviewer diagrams the sequence of events on a timeline  Identify critical points/”decision points”  Try to note sequence and duration of events, actions, perceptions,

thoughts, decisions

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

12  

Sweep 3: Deepening

  Goal: Get inside the expert’s head: ”[W]hat did they know, when did they know it, how did they know, and what did they do with what they knew?” (Crandall et al., 2006)

  Based on the timeline probe critical points for the participant’s  Perceptions  Expectations  Goals  Judgments  Confusions, uncertainties, concerns  Options  Information needed and used

2011-04-18 carolina.wannheden@ki.se

Sweep 4: ”What if” queries

  Goal: Illuminate expert-novice differences and potential vulnerabilities for error in the domain

  The interviewer poses hypotheticals about the event  What if a novice had been at charge?  What if [key feature] had been different?  What training might have been an advantage?  What knowledge, information, tools/technologies could have

helped?

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

13  

Discuss in your groups…

  Would the Critical Decision Method be an appropriate method to elicit knowledge for the CDSS you intend to develop?

  Would it be a feasible method?

2011-04-18 carolina.wannheden@ki.se

A final quote

”For now, […], it is the direct interaction among experts, and between experts and knowledge engineers, that will serve a crucial role in assuring the development of high quality and accepted knowledge bases that in turn enable the development and effective use of decision support systems.” (Greenes, 2006)

2011-04-18 carolina.wannheden@ki.se

11-­‐04-­‐18  

14  

WW DD

2011-04-18 carolina.wannheden@ki.se