COMPLEX PROBLEMS CLASS 3

1
COMPLEX PROBLEMSCLASS 3

• Heuristics to the Rescue
• Alternatives to Purely Analytical
• Problem-Solving Methods

2
Heuristics as a Problem

• Cognitive Biases -- Implicit Heuristics
• Most of the early work on heuristics focused on
  showing poorly understood use of heuristics when
  thinking/using analytical methods (Kahneman
  Tversky) --
• Strong Priors
• Unwarranted Analogies
• Representative Bias
• Myopia
• Control ...

3
Heuristics as a Problem-Solving Tool

• Heuristic
• helping to discover or learn a method of
  education learning or computer programming in
  which the pupil or machine proceeds along
  empirical lines, using rules of thumb, to find
  solutions or answers (New World Dictionary, 2nd
  Ed.)
• In this broad positive sense, heuristics are
  problem- solving methods using mental shortcuts,
  trial error, ...
• Rules of Thumb
• Empirical Searches exploratory experiments
  (Trial Error)
• Intuition experience

4
Non-analytic Based Disciplines

• Does not imply lack of intelligent thought or
  complete absence of analytical methods, just not
  the core of problem solving
• Significant Heuristic Content or Method
• Biomedicine
• Cognitive Psychology
• Computer Science, Engineering, Stats, Math ...

5
Contrasting Analytical Heuristic Methods

• Math Example
• Find point where y 2x2 - 10x reaches a
  minimum
• Analytical Find derivative, set equal to zero
  and solve for x
• (dy/dx 2x - 10) (0 2x - 10) (x 5)
• Heuristic Numerical Search (trial error)
  plug-in values for x until you find the minimum
  point
• in case above, analytical approach more
  efficient
• if equation not easily solved with analytics,
  numerical search more efficient or may be only
  possible

6
Heuristics in Analytics

• Prior Class Analytical Toolkit
• Analogy
• Solving in parts
• Backward-Forward
• Transforming into known problem
• Generalizing from specific solution
• Although now integral part of analytical
  problem-solving techniques, these are really
  rules of thumb (heuristics) that people tried
  over time they became widely known because they
  worked

7
Experimental Evidence on Heuristics

• Bargaining Ultimatum Game
• Fixed sum to split between Player A Player B
  A makes offer, if B rejects, game ends-no deal
  if B accepts, deal made
• Analytics A offer minimum, B accept
• Experiments
• Few offers below 7525 split, if made, rejected
• Reasons fear rejection, fairness so use
  heuristics (intuition )
• Extensions time limits,multiple rounds,
• people who quickly agreed on 5050 did the best
• people who tried for very uneven results or
  bickered over small differences did the worst
• Bottom Line?
• Where norms are involved that are not easily
  model in analytics, using intuition can improve
  on pure analytics

8
Experimental Economics Evidence on Heuristics

• Focal Points -- rule of thumb (heuristic)
  solutions to difficult strategic decisions
• Divide the cities (location game) Harvard
  Stanford MBA student pairs separately choose from
  list of cities with certain limitations score
  bigger when less overlap
• Results?
• Geographic (East-West) focal points
• Where to Meet in NYC?
• Thomas Schelling experiments found GST focal
  point
• Bottom Line?
• Intuition can be a useful heuristic

9
Biomedical Examples

• Rules of Thumb Intuition
• Patient presents with sore throat upper
  respiratory symptoms
• common cold virus
• Infant presents with intestinal discomfort
• gas
• Patient presents with shortness of breath and
  fatigue
• asthma
• Search
• Patient presents with fever severe sore throat,
  fatigue -- no other
• Rapid Strep Test
• Symptoms recur frequently -- overnight strep
  culture
• Lack of resolution -- blood tests (C-reactive
  protein white blood cells )
• Start from low cost/most common and proceed

10
Example from Everyday Decisions

• Heuristics that you employ or have seen others
  employ in everyday personal decisions
• Rules of Thumb?
• Searches?
• Take 15 minutes and write down list of heuristics
  from both personal and business experiences
  rank by the ones that work well and not so well
  indicate why
• In general, these kinds of examples from are
  often called Satisficing

11
Limits of Intuition

• Fooled by Randomness (Taleb, 2002) Experiments
  show that people using simple observation
  (heuristic), repeatedly see patterns to outcomes
  where none exist

12
Robert Lucas on Limits of Simple Searches
Observations

• Since the mid 1980s, companies like Microsoft,
  FedEx, Staples, MCI and many others have shown
  tremendous growth in earnings, market share,
  employment and other performance measures. With
  such a range of experience, why do we need
  theoretical models? Why cant a company just send
  a fact-finding team to Staples, find out the
  strategies and structures which made them
  successful, and then go home and get their own
  company to do the same? This sounds easy enough,
  but it is not really operational ... Firms are
  just too complex -- there are too many things
  going on at once -- for getting all the facts to
  be either possible or useful. Faced with so much
  data, an observer who is unequipped with a theory
  sees what he wants to see, or what the successful
  company or management guru wants to show him.
  One needs some principles for deciding which
  facts are central and which are peripheral. This
  is exactly the purpose of to isolate some very
  limited aspects of a situation and focus on them
  to the exclusion of all others ... We need to
  make some hard choices about what to emphasize
  and what to leave out before we can think in an
  organized way at all.How do economies or
  companies succeed?
• Adapted from Robert Lucas, 1994 Nobel Prize
  Winner to European Econometrics Society
  (Extension of science philosopher, Karl Popper)

13
Limits of Heuristics Examples

• Biomedical
• Jerome Groopman -- Harvard Hematologist-Oncologist
  Second Opinions Stories of Intuition and Choice
  from the Changing World of Medicine (2001)
• Pediatricians rule of thumb gt Infant has gas
• Reality gt Obstructed Bowel
• Primary Physician rule of thumb gt Woman has
  asthma
• Reality gt Leukemia
• Production
• Carnegie-Mellon Bicycle Production Example

14
General Lessons about Using Heursitics

• Heuristics Most Successful Where
• Outcomes conform to typical situations
• Heuristics fail where nuances present
• Causal relationships simple (e.g. virus-disease)
• Heuristic approach in biomedicine not as
  successful in multi-factor problems, e.g.
  Neurological problems
• Problem of a nature so that computational power
  can overwhelm the search problem by brute force,
• e.g. Human Genome project finding a numerical
  solution through searches
• Problem is time-sensitive
• Problem beyond analytical limits

15
The Tradeoffs analytics v. heuristics

• Analytics -- High Cost/Low Error
• Precise, Logical,
• Costly in terms of time/mental demands
  flexibility
• Heuristics -- Low Cost/High Error
• Low cost in terms of time/mental demands,
  flexibility
• Difficult to assess Subject to unclear biases
• CS Heuristics are bug ridden by definition. If
  they werent, they would be algorithms

16
Analytical Heuristics an oxymoron?

• Methods different but the same?
• rules of thumb intuitive searches sometimes
  imitate analytical results (Day, AER)
• Analytical Heuristics
• Not all search random are led by intuition alone
• Search can be guided by analytics prior
  knowledge
• Wright Brothers entrepreneurship

17
Development of Flight Bradshaw Lienert (1991)
18
WRIGHT BROTHERS ENTREPRENUERSHIP

• One meaning of entrepreneurship innovating and
  improving through combining or mutating products,
  processes
• Wright Brothers (See Related Websites First
  Flight)
• Did not just go through trial error or use just
  intuition
• Serious research on past efforts
• Analytical reasoning regarding physics
• Then Search through Experimentation in wind
  tunnels and full-scale

19
Critical Lessons

• Effective use of heuristics will make you a
  better manager
• Overuse of heuristics will create mistakes --
  sometimes devastating mistakes
• Employ heuristics when best suited to the
  situation -- not as a crutch to alleviate the
  need to think analytically
• What is the decision environment
• Time sensitivity, analytical tractability and
  completeness, search power, analytics guiding
  heuristics possible
• Remember the implicit biases apply to heuristics
  too!

20
Mini-Assignment

• Identify and be able to explain 2 examples of the
  use of heuristics to solve problems in a
  workplace setting involving rules of thumb,
  simple search, or the overlap of analytics
  heuristics. Evaluate how well the heuristic
  seems to work and the reasons it does or does not
  seem to work very well.