dynamic foraging model for human behavior on the internet (working title) bjarne berg
TRANSCRIPT
2
Agenda
Introduction and Background
Optimal Foraging Theory (OFT)
Foraging Models
Economic perspective (EMH and REH)
Human Factors
Proposed Framework
Question and Answers
3
Background
Despite extensive research in the human evolution since 1854 when Charles Darwin published his thesis on the Origin of the Species, the motivation and the behavior associated with foraging and thereby the natural selection through the survival of the best foragers, was not well researched for over one hundred years.
This changed in 1966 when researchers such as Emlem published his work on the foraging behavior of birds and by the publication the same year of MacArthur & Pianka’s work on the individual’s selection process of foraging areas. Over the next thirty years this research gave rise to a large field known as Optimal Foraging Theory (OFT) that has been the foundation in a variety of biological and zoology studies.
Only in the last ten years OFT has been extended into the field of information technology and search algorithms (Sugawara and Watanabe, 2002; Pui and Huosheng, 2002).
4
Background
At the same time, there has been an increased interest in the last decade of extending OFT into a better understanding of human behavior on the internet through intelligent foraging agents (Jiming et. al., 2004) and through extensions to social behaviors of foraging agents (Andrews, 2007).
Most of this research has focused on optimizing the algorithms of robots or intelligent agents that can, on behalf of the human, scan vast amounts of information to find specific items.
5
Agenda
Introduction and Background
Optimal Foraging Theory (OFT)
Foraging Models
Economic perspective (EMH and REH)
Human Factors
Proposed Framework
Question and Answers
6
The Optimal Foraging Theory (OFT)
In 1966 the field known as Optimal Foraging Theory (OFT) was established through the publication of Emlen’s article on foraging behavior of birds and by MacArthur & Pianka’s work on optimization models the same year.
In general, the models established over the next ten years focused on four core areas that became known as elements of a micro-ecological theory. These areas include
1) What to eat (optimal diet). 2) Where to eat (optimal patch choice). 3) Optimal allocation to each patch (time). 4) Optimal patterns and speed of movements.
Combined as a whole, the micro-ecological theory forms the platform for macro-ecological theory which has far reaching implications
7
The Optimal Foraging Theory (OFT)
Natural Selection – The optimal should already be here
Cost- Benefit and minimal benefit requirements
Optimal Patch Choice (OPC)– scholastic models (bird’s patch selection)
Committed exclusions and logical progressions
Sub-optimal foraging – social and cultural constraints
Compression Hypothesis - as the number of competing species increase, a reduction in the patches used occur & the range of items consumed remains constant or only slightly increase .
Specialization – Koala Bears (increased food abundance leads to greater food preference)
8
Agenda
Introduction and Background
Optimal Foraging Theory (OFT)
Foraging Models
Economic perspective (EMH and REH)
Human Factors
Proposed Framework
Question and Answers
9
Foraging Models
Optimal Diet Theory (ODT) - also includes advanced mechanisms for gradual shifts in item acquisitions when preferable items (high net benefit) exists or becomes more abundant.
Optimal Time Allocation (OTA) and Marginal Value Theorem (Chernov, 1976),
and Surrender Time
Evolutionary foraging algorithmsThe simplest approach to account for the dynamically changing environment has been to introduce uncertainty/variability into the approach and rebuild the new optimal search patterns and speed of movements each time a foraging event on a patch, or set of patches, are completed (Yang & Yao, 2005). This was a focus area in the OFT research field in the late 1990s and 00s. The number of recent models proposed using this approach are numerous (Branke, 2002; Jin & Branke, 2005; Tin´os & Yang, 2005).
10
Evolutionary Foraging Algorithms (EA)
1. Bacterial foraging algorithms (BFA) and Dynamic BFA (DBFA)
2. Group foraging theory and diversity in Evolutionary algorithms
3. Dynamic and Memory enhanced foraging algorithms (E. Coli)
4. Thermodynamical Genetic Algorithm (TDGA)
11
Agenda
Introduction and Background
Optimal Foraging Theory (OFT)
Foraging Models
Economic perspective (EMH and REH)
Human Factors
Proposed Framework
Question and Answers
12
EMH and REH
1. Efficient Market Hypothesis
2. Rational Expectance Hypothesis
3. Price Dispersion and 2-step models
13
Agenda
Introduction and Background
Optimal Foraging Theory (OFT)
Foraging Models
Economic perspective (EMH and REH)
Human Factors
Proposed Framework
Question and Answers
14
Human Factors
Technology
Basic Technology• Input, output devices• Interaction styles• End-user computing• Org. computing
Advanced Technology• Information visualization• Perceptual/attentive/embodied/
multi-modal/portable/wearable/ implant/personalization
• Persuasive computing• Affective computing
Task / job• Task goals• Task character• Task complexity
Human
Demographics• Gender, age, culture• Comp. experience• Education
Cognition• Cognitive style• Perception• Attention• Memory• Knowledge• Learning• Error• Distributed cognition
Physical/Motor• Motor control• Comfort
Emotion & Motivation• Affectivity• Affective state• Mood/feeling• Emotion• Intrinsic motivation• Extrinsic motivation
Context
Global Context• National culture• Norms• Universal accessibility
Social Context• Privacy•Trust• Ethics• Norms
Org. Context• Org. goals• Org. culture & norms• Policy & procedures• Management support
Group Context• Group goals• Group norms
De
sig
n
Us
e
Imp
ac
t
Source: Zhang and Li Review of Intellectual Development of HCI Research, 2005 (13 yrs of articles in 7 top journal – 348 articles)
15
Agenda
Introduction and Background
Optimal Foraging Theory (OFT)
Foraging Models
Economic perspective (EMH and REH)
Human Factors
Proposed Framework
Question and Answers
16
Constructs
The task of the seller is to minimize the consumer surplus, while the task of the consumer is to maximize it. It is important to note that if there are no consumer surplus, the sale cannot occur (consumers would be unwilling to proceed).
Therefore some consumer surplus has to exist, however marginal. In a foraging model an implicit equilibrium should exists between the consumer price, the foraging costs and the foraging surplus on one side and the optimal (best price available) and the incremental foraging costs of locating this best price.
FCFCOPTSFCCP
Or simplified: FCOPTSCP
CP = Cost of item (paid) FC = Foraging costs S = Foraging surplus OPT = Optimal price available (all patches) ∆FC = Additional foraging costs required to locate optimal price
17
Constructs
FCCPFCOPTS
FCFCOPTSFCCP
CP = Cost of item (paid) FC = Foraging costs S = Foraging surplus OPT = Optimal price available (all patches) ∆FC = Additional foraging costs required to locate optimal price
18
Time Constructs
ID Variable Definition Calculation Example
TA Time to Access
The time it takes before the patch becomes accessible
Measured as time from last action until the site is available
Load time of www.delta.com airline web site
TO Time to Orient
Time it takes from entering a patch to become informed of its purpose and content
Measured from first availability of the site to the next action is undertaken
Time from the web site is loaded until customer undertakes an action
TE Time Enter
The time it takes to enter all the required search criteria to search products or services at a patch.
Measured as time from beginning entering a search at a patch until action is submitted.
Time it takes to enter a search for a flight between two cities at a given day and for a coach ticket.
TF Time to Find
Time to execute search at a patch
Measured from the search action is submitted until the complete set of options are available
Time it takes for Delta's web site to execute the search and present the results (i.e. 10 possible flights).
TR Time to Review
Time it takes to review the items available at a patch after search has been completed.
Measured from the time a result set has been presented until next non review action is taken
Time it takes a customer to review the 10 flights and pick the best option, try another search, or leave the website.
TZ Time to Acquire
Time it takes to take possession of an item
Measured as the time an item has been identified as the solution, until ownership has transferred
The time it takes to select the flight, enter the passenger name, credit card and other data to the purchase confirmation is received.
19
Foraging Costs and Incremental Foraging Costs
p
iiiii tztstotaEFC
1
)(
E = cost per unit of time p = number of patches accessed ta = Time to access patch to = Time to orient at patch ts = Time to search tz = Time to acquire item
p
piiiii tztstotaEFC )(
20
Foraging Search Costs
xq
jjxxx
s
x
trtftets1
,1
)(
s = Number of searches at a patch te = Time to enter search tf = Time to find (execute search) q = Number of items returned for review tr = Time to review each item
21
Overall model (work in progress)
n
kk
p
i
s
x
q
jjxikxikxikikikkkk tztrtftetotaECPFCOPTS
k ik xik
1 1 1 1,,,,,,,,,
, ,,
Experiment identifier
(id)
Survey number
(no)
Optimal price
(OPT )
Price paid (CP)
Patch number
(p)
Patch name (url)
Patch departure
(buy/surrender)
Best price located
patch/search(bqs)
Time to access
(ta)
Time to orient (to)
Search number
(s)
Time to enter (te)
Time to search
(tf)
Number of items
reviewed(q)
Total time to review
S(tr)
Average time to review
(mean tr = S(tr)/q)
Time to acquire
(tz)
No Nm Cd No
1 13 $297 $397 1 delta.com s $384 11 5 1 14 23 22 128 5.82
1 delta.com s $469 11 5 2 13 25 26 105 4.04
1 delta.com s $498 11 5 3 10 26 17 159 5.47
2 Usair.com s $440 8 14 1 16 17 135 139 1.03
2 Usair.com s $635 8 14 2 15 19 122 128 1.05
2 Usair.com s $402 8 14 3 16 19 164 240 0.24
3 Orbiz.com b $397 9 13 1 9 14 14 131 9.36 189
2 13 $165 $165 1 hertz.com s $171 15 5 1 13 22 21 102 4.85
1 hertz.com s $178 15 5 2 12 24 25 84 3.37
2 avis.com s $171 8 13 1 10 25 16 74 4.56
2 avis.com s $209 8 13 2 15 9 130 111 0.86
2 avis.com s $168 8 13 3 14 9 117 102 0.87
3 national.com s $211 9 6 1 15 10 157 32 0.20
3 national.com b $165 9 6 2 9 6 6 65 10.83 206
Seconds SecondsNo AmtAmt
22
Example - calculations
Experiment identifier
(id)
Survey number
(no)
Optimal price
(OPT )
Price paid (CP)
Foraging surplus
(s )
Patch number
(p)
Patch name (url)
Patch departure
(buy/surrender)
Surrender time
(seconds)
Incremental foraging
time( ft)
Cumulative foraging
timeS(ftp,s)
Incremental foraging
costs (FC)
Cumulative foraging
costs S(FC)
Best price located
patch/search(bqs)
Incremental foraging surplus
s
No Nm Cd
1 13 $297 $397 -$112 1 delta.com s 181 181 $1.51 $1.51 $384 -$85.5
1 delta.com s 143 324 $1.19 $2.70 $469 -$86.2
1 delta.com s 519 195 519 $1.63 $4.33 $498 -$30.6
2 Usair.com s 172 691 $1.43 $5.76 $440 $56.6
2 Usair.com s 162 853 $1.35 $7.11 $635 -$196.4
2 Usair.com s 609 275 1128 $2.29 $9.40 $402 $230.7
3 Orbiz.com b - 154 1282 $1.28 $10.68 $397 $3.7
2 13 $165 $165 -$8 1 hertz.com s 158 158 $1.31 $1.31 $171 $127.3
1 hertz.com s 278 120 278 $1.00 $2.32 $178 -$8.0
2 avis.com s 109 387 $.91 $3.23 $171 $6.1
2 avis.com s 136 523 $1.13 $4.36 $209 -$39.1
2 avis.com s 126 649 $1.05 $5.40 $168 $40.0
3 national.com s 428 57 706 $.48 $5.88 $211 -$43.5
3 national.com b - 80 786 $.66 $6.55 $165 $45.3
No Seconds AmtAmt
23
Some hypothesis
HYPOTHESIS 1: Surrender events increases foraging surplus of participants
(surrender benefits) in electronic commerce.
HYPOTHESIS 2: As the number of available e-commerce sites for a given
(specialization) product increases, usage consolidates to a few sites.
HYPOTHESIS 3: A high number of items returned in a given search has a
(information overload) negative influence on the foraging surplus realized by
actors in an e-commerce marketplace and the impact is not
uniform for all participants.
HYPOTHESIS 4: A negative relationship exists between foraging surplus and
(item availability) a low number of items returned by a search.
24
Some hypothesis
HYPOTHESIS 5: The process of site exhaustion (SE) reduces radical changes
(site exhaustion) in patch choices and is inversely related to previous
experience.
HYPOTHESIS 6a: Actors that exhibits a moderate propensity to explore
(exploration) increase their foraging surplus.
HYPOTHESIS 6b: A high propensity to explore is negatively related to
(exploration) foraging surplus (s).
HYPOTHESIS 6c: Age is a factor in the actor’s propensity to explore.
(age - exploration)
HYPOTHESIS 6d: The propensity to explore is directly related to previous
(experience - exploration) experience.