This house believes that A

1
(No Transcript)
2
(No Transcript)
3
This house believes that AI is RIPOpposing the
motionNick Woolley, Information Specialist,
Kings College LondonNovember 2009
4
Preaching to the converted?

• Do our old axioms still hold true?
• Our are cherished constants still relevant given
  the Web and Web 2.0 / 3.0 / number of your
  choice?
• In theory there shouldnt be much to debate
• - recall, precision
• - exhaustivity, specificity
• - information needs
• - information seeking behaviour
• - ontology, taxonomy

4
5
AI perspectives and issues

• Technology
• Findability
• Information overload
• User requirements
• Value for money (VFM)

5
6
AI technology

• Human and automatic, manual and algorithm
• Expert and amateur
• Controlled vocabulary and folksonomy
• Full text (web) and selective index
• Be a lumper and think metadata
• Of course, not all metadata is equal (or really
  meta!)

7
Google AI

• Google is AI, Google Scholar is AI, so is
  Google Books
• Just not very good AI
• Bad metadata, ghost authors for citation data
• See - Peter Jascow Google Scholars Ghost
  Authors, Lost Authors, and Other Problems
  Library Journal, 9/24/2009
• http//www.libraryjournal.com/article/CA6698580.ht
  ml?rid1105906703sourcetitle
• From what we can figure out!

7
8
Findability

• Peter Moville's Ambient Findability (2005)
• Ambient findability describes a fast emerging
  world where we can find anyone or anything from
  anywhere at anytime. (p.6)
• ..ambient findability is less about the
  computer than the complex interactions between
  humans and information.(p.13)
• The better the metadata the greater the
  findability

8
9
User requirements

• Still display Anomalous States of Knowledge (ASK)
• Digital natives and similar myths
• Exhibit the cultural norm of instant
  gratification
• Want and expect full text now
• Paying more for the service
• Disintermediation is still a relevant concept,
  especially when ignorance is bliss (not Bliss)

9
10
Ignorance is bliss (theory)

• Mooers law (1959) - original context was
  contradictory principle of behaviour not
  usability
• .. many people may not want information, and
  that they will avoid using a system precisely
  because it gives them information
• Where an information retrieval system tends not
  to be used, a more capable information retrieval
  system may tend to be used even less
• See Austin, B. 2001 http//spot.colorado.edu/norc
  irc/Mooers.html

10
11
Ignorance is bliss (real world example)

• The Rationally Ignorant Consumer Hypothesis
• Media and markets, imperfect and biased
  information provision
• People choose to be RICs
• McCluskey, J.J. Swinnen, J.F.M. 2004,
  "Political economy of the media and consumer
  perceptions of biotechnology", American Journal
  Agricultural Economics, 86 (5) 1230-1237.

11
12
Information Overload

• The haystack is getting bigger
• Negates labour-saving technology
• Less signal more noise, deflates precision and
  recall
• More RICs and ASKs get more anomalous?
• Futility Point Criterion decreases as of total
  relevant records
• Information overload opposes findability.
• AI becomes more important.

12
13
Wisdom of crowds?

• End user generated metadata for free!
• Do folksonomies enhance findability?
• Are tags and related activity a valid form of
  AI?
• Can mob indexing solve information overload (or
  just add to it)?
• Personal versus shared worlds
• Nothing is free!

13
14
Losing the Long Tail?

• The Long Tail thrives online
• Web 2.0 crowd AI is still crude
• Little exhaustive or specific indexing so low
  recall
• Reinforcing homogeneity, dampening serendipity
• Even with good AI not all documents are equally
  findable, but a degree of parity helps
• Can you conflate popularity with value?
• How do Tag clouds work?

14
15
LibraryThing
0 tags v ? (a lot) Can you guess the book on the
right?
15
16
One stop shops?

• Single interface and aggregation
• Zipf and Mann - principle of least effort
• Familiarity
• Joins up institution resources
• Added value not replacements

16
17
VFM

• How do you value findability?
• Good AI increases chance of uniting researcher
  with information they need
• This can translate into revenue far exceeding the
  cost of the service
• AI contributes to the success of your institution

17
18
VFM

• Discovery or Access
• Database expenditure lt ejournals expenditure
• When budgets bite why go after the smaller slice?
• Controversial, e.g.
• RIN ejournals report April 2009
• Perceived and / or real immediate need for
  information
• Which would you choose Pubget or EndNote X3?

18
19
VFM

• Back of an envelope calculation
• Cost of database subscription pa a
• Cost of staff time searching and supporting
  databases pa b
• Cost of staff time searching Google pa c
• a b lt c
• False economy and decline in standards?

19
20
VFM

• Look at the evidence
• - Do your users use your subscription databases?
• - Which groups benefit most?
• - Do researchers rely on your databases?
• - Are you asked to provide details for grant
  applications?
• - Do you support systematic reviews?
• Imagine you have no paid AI. What happens next?

20
21
Challenges

• Education, education, education
• Information literacy, information skills
• Advocacy and marketing
• Identifying and demonstrating value
• Collaboration between publishers and information
  services

21
22
Summary

• AI carries on evolving 90s new automated
  methods, 00s mob indexing, new interfaces and
  aggregation
• Whatever your bottom line (findability, VFM) AI
  is essential
• Abandoning AI can only be a false economy
• We need to face up to the challenges
• Integrating and joining up with new AI is vital,
  nothing is mutually exclusive.

22
23

• AI is alive and kicking!

23