Oop'Narf and Up The Junction: Capturing the Vernacular

1
Oop'Narf and Up The Junction Capturing the
Vernacular

• Andy Evans

2
Thanks

• Steve Carver (Leeds Uni)
• Richard Kingston (Manchester Uni)
• Tim Waters (Bradford Council)
• Chris Jones (Leeds Uni)
• Kevin Cressy (City Uni)
• Without whom

3
Whats it all about?

• Work on democracy
• How people relate to the world
• Capturing vernacular geography
• Using vernacular geography

4
Whats wrong with the political process?

• 2001 average election turnout 59.4
• Nottingham East 45.5
• Nottingham South 50.11
• Nottingham North 46.74
• Most of the missing are in the 18-34 age bracket.
• This is 12 lower than anything since WWI.
• Turnout for local elections 30 or less
• Yet, people arent content
• We have 1.5 Million people marching through
  London.
• Support for local causes seems to be on the up.

5
Public participation and apathy

• People may feel
• The system doesnt cover the things they are
  interested in.
• All politicians are the same the system doesnt
  have the breadth of representation to express
  their views.
• The politicians who hold their views on one
  thing, may not on others.

• When a by-election (Leeds Central) has a turnout
  of 20, representational democracy has failed.
• Yet participatory democracy is increasing.

6
The dangers of participation

• Tends to be dominated by a few.
• Agenda can often be over-constrained.
• Tends to be difficult to get involved.
• E.g. Drive to meetings.
• Information often difficult.
• But, people do want to be involved.

7
GIS and the Web

• GIS can help
• Frame questions in relationship to where people
  live.
• Allow people to explore scenarios and data.
• The web can help
• Access better than meetings.
• Less domination by small groups.
• More time for consideration.
• Easier data access (for public) and analysis (for
  decision makers).

8
Example Virtual Slaithwaite

• Planning for Real.
• Took over village fair as well.
• Allowed input of problems attached to locations.
• Easy analysis Built community understanding of
  level of concern about locations and issues.

9
Example Multi-Criteria Evaluation

• Where is Wild in Britain?
• Where should the Yorkshire Dales have more trees?
• Where should we dispose of Nuclear Waste?
• Rank a number of factors and constraints.
• Allowed analysis of how people respond to data
  and opportunities for change.

10
Fear of the great unwashed

• Analysed factors important to people in nuclear
  waste sites before and after they saw
• The geography
• Where their best sites were relative to their
  homes.
• Geography did effect the chosen location/factors.
• But didnt effect how close to their homes.
• People made rational choices based on their
  preferences.

11
Deeper understanding of risk and location

• Current project with Wakefield Council.
• Allows people to zoom into a map and comment on a
  problem.
• Burnt out car, graffitti, dead animals, noise
  etc.
• We hope it will show
• the way people navigate data.
• the scale at which people understand different
  problems.

12
Problems

• Access 53 of homes online
• strong age skew.
• 7 thought too expensive.
• Taking systems to the people.
• More generally we need to balance hard to
  understand data vs. completeness.

13
How people relate to the world

• Work on democracy
• How people relate to the world
• Capturing vernacular geography
• Using vernacular geography

14
Normal Human Beings?

• The General Populous are darned fools
• They use geographical terms they cant define.
• They mix up their attribute datasets.
• They can rarely put anything precisely on a map.
• Why, oh why, cant The General Populous use
  geographical coordinates and specific data layers
  like Normal Human Beings?

15
Vernacular Geography
Locational Loaded
Uptown Our village The shops Snowdon The East End Down by the gasworks Up the Junction Across the river The Fens Down by me Nans N.B.Places and relations Dangerous end of town High crime area Ugly bit of the suburbs Dodgy area around the station The Ghetto The simply delightful area around the park Commutersville
16
Vernacular geography

• When asked, for example, to define and explain
  areas where they are afraid to walk in the dark
• The datasets people use are continuous and
  discrete, at differing scales, historical,
  architectural, and mythological.
• The resultant areas linguistically ambiguous.
• May be bound by prominent landscape features for
  convenience, but are more usually diffuse.
• Often have different levels of intensity within
  the areas.

17
Vernacular geography is good.

• Evolved to make things easy to remember and
  discuss.
• Gives us geographical references that include
  associated environmental, socio-economic, and
  architectural data.
• He lives in the grim area by the docks
• Im going down to the shops
• Gives us a connected socio-linguistic community
  with shared understandings (and prejudices).
• A poor little baby child is born In the ghetto
  
• This is a local shop, for local people

18
Vernacular geography is important.

• Represents psychogeographical areas in which we
  constrain our activities.
• I wouldnt walk through the rough bit of town at
  night
• Conveys to our socio-linguistic community that
  this constraint should be added to their shared
  knowledge and acted upon.
• Thats a pretty high crime area
• This private and shared geography influences
  billions of people every day.
• But its hard to tie directly to objective data
  so we can use it to make policy or scientific
  decisions.

19
Capturing vernacular geography

• Work on democracy
• How people relate to the world
• Capturing vernacular geography
• Using vernacular geography

20
Are fuzzy boundaries useful?

• Fuzzy boundaries occur because of
• Continuousness (Ontic vagueness)
• When the datas not discrete.
• Aggregation (Prototyping)
• Where discrete boundaries represent the average
  location of continuous or discrete variables
  binned together for descriptive convenience,
  usually categorized by comparison with a
  prototype.
• Averaging (Scale dependent vagueness)
• Where discrete boundaries average a single time
  or scale varying geographical boundary.
• Imprecision (Epistemological vagueness)
• Where we cannot know a boundary because we cant
  measure it accurately enough.
• Ambiguity (Semantic vagueness)
• Where boundaries are tied to linguistic factors.

21
Tools in three interfaces

• Weve been developing a set of tools to capture
  fuzziness in a GIS.
• Input
• A spraycan interface for a online GIS, that
  allows comment attributes to be attached.
• Administration
• For decompression and combination.
• Query
• A way of representing all users data and
  searching for the comments in order of users
  perceived importance.

22
Input GUI

• Spraycan of different sizes.
• Attribute information box.
• Send button.

23
Type of Spraycan

• Continuous vs. dot
• Gaussian vs. constant
• Users preferred constant fills.
• Easier to use where boundaries a mix of sharp and
  fuzzy.
• Dots are easier for beginners.
• Easier to grasp the intensity the spray is at in
  white space areas.
• Harder to convert into a density map.

24
Output GUI

• Click on map of combined areas.
• Comments of the people who weighted that area as
  most important float to the top.

25
Its not a perfect world

• Transferring data across the net.
• Combining and searching many user responses.
• Need to balance the accuracy of our
  representation with the technical difficulties.

26
The compression
Waveform compression algorithms might improve on
this.
Tests suggest a typical compression rate is two
orders of magnitude. For example, a combined
image and data object of 859Kb was compressed to
67Kb just using the GZip algorithm, and further
compressed to 14Kb with the addition of the
shrinking process.
27
Technicalities

• User tests suggested a 9x9 pixel averaging kernel
  best represented the areas users had drawn using
  the dots.
• Tests suggested this could be shrunk to 5 times
  the size and re-inflated without users noticing a
  significant change in the image.

28
Recent developments

• New system to capture these areas in Arc.
• New system to allow you to use a pencil in Arc
  to draw boundaries.
• New server-side system which speeds up
  implementation and scalability.
• http//www.ccg.leeds.ac.uk/software/tagger

29
Capturing vernacular geography

• Work on democracy
• How people relate to the world
• Capturing vernacular geography
• Using vernacular geography

30
Capturing High Crime Areas

• 2001/2002 British Crime Survey people have a
  higher fear of crimes than real victimhood.
• Believe crime rates are increasing, most actually
  falling.
• The fear of crime has a significant impact on
  peoples lives
• 7 go out less than once a month because of the
  fear of crime.
• 29 of respondents say they didnt go out alone
  at night.
• 6 said fear of crime had a great effect on
  their quality of life.
• 31 said it had a moderate effect.
• Concern about crime therefore represents a
  significant influence on many peoples lives.

31
Case study Crime in Leeds

• Where do people think are the High Crime areas
  in Leeds?
• 50 users drawn from various socioeconomic levels
  from all over the area.
• Blue are areas safer than thought, red less
  safe.
• People could see how others felt about areas.

32
First we need to understand the data

• There are clear problems in this (toy) analysis.
• How can such entities be compared with
  traditional scientific data?
• What kinds of algebra can be performed on such
  data, alone and in combination with other
  datasets?
• How do we deal with neighbourhood influences both
  within and between fuzzy neighbourhoods.

33
How might we handle it Fuzzy logic

• The notion that something is more or less
  something can be handled by fuzzy logic.
• Scientists love it so much
• Fuzzy theory is wrong, wrong, and pernicious.
  What we need is more logical thinking, not less.
  The danger of fuzzy logic is that it will
  encourage the sort of imprecise thinking that has
  brought us so much trouble. Fuzzy logic is the
  cocaine of science.
• Prof William Kahan

34
Fuzzy rules

• Users sprays represent membership values for
  each point of a fuzzy set, e.g. CRIMEFEAR.
• We can then build up rules
• if (CRIMEFEAR is HIGH) and
• (REALCRIME gt average) then
• INVESTMENT is HIGH
• Pros Gives you some degree something is true.
• Cons Hard to know how to use. E.g. Union of sets.

35
How might we handle it

• Supervaluation logic
• Assumes all vagueness is linguistic.
• Attaches the same term to different distinct
  boundaries.
• i.e. We draw multiple examples of definite
  boundaries.
• Analysis examples
• Something is super-true if it is true for all
  definitions.
• Something is definitely possible if it is true
  for one definition.
• Pros Gives definite maybes.
• Cons Assumes definite boundaries can be drawn.

36
Mereotopological calculi

• Areas defined like fried-eggs.
• You can make definite statements about some bits,
  and not about others.
• Pros Useful for qualitative relationships A is
  next to B.
• Cons No real notion of complex gradients / 2nd
  order vagueness.

37
Problems

• In fact, theres nothing to stop us breaking our
  data down in all these ways, once it is in.
• More pressing problems
• What do the numbers represent?
• What is the maximum in this situation?
• How do we combine data from one person?
• Contrasting e.g. levels of RURAL vs. URBAN?
• Different categories e.g. MOUNTAIN vs. FOREST?
• How do we combine data from multiple people?
• Variety of connectors in fuzzy logic and decision
  support theory, but often seem arbitrary.

38
Problems

• Can we disaggregate into different types of
  fuzziness/causes?
• Do we need to?
• Is data fit-for-purpose if not collected for a
  specific use?
• Does supervaluation etc. sets allow us to get
  round this?
• How do we compare this with real data?
• Confusion / Entropy indexes?
• Could treat it as a set of beliefs (or beliefs
  about memberships)
• Doxastic Modal logics, Evidence set theory.
• Could treat it as a probability and use Bayesian
  statistics
• Though this may mis-represent one person, it may
  model combinations of areas.

39
Problems

• How do we cope with resolution?
• We may collect at one scale and use at another.
• Can we quantify or predict the erroneousness of
  scale changes?

40
Example analyses

• How does fear of crime vary with
• personal victimhood?
• media exposure?
• conditions (summer vs. winter)?
• Current models based on aspatial demographic,
  psychological and temporal factors only accounted
  for 1/3 aspatial fear levels.

41
More generally

• Policy Where should we invest to improve
  perceptions?
• Psychology What is the relationships between
  things in the real world and perceived areas?
• Is there a predictable relationship?
• Are they at the same place?
• Does perception of some things have a wider
  geographical spread than others?
• How to people get an understanding of areas?

42
Future

• Most work has focused on
• Storing data so qualitative spatial relationships
  can be generated (next to, touching, within,
  etc.).
• Capturing quantitative spatial relationships
  using fuzzy logic (close to, far from).
• How often are these used in policy making?
• Is it better to concentrate on how we relate this
  data to the real world and similar datasets?
• Vernacular geography is vastly more complex
  though.
• All lines are fuzzy (measurement / labels) weve
  just hidden it.

43
Further information

• www.geog.leeds.ac.uk/people/a.evans/
• www.ccg.leeds.ac.uk/democracy/
• www.ccg.leeds.ac.uk/software/tagger/