Dubois E., Gray P., Nigay L. (Eds.) The Engineering of Mixed Reality Systems

Подождите немного. Документ загружается.

86 A. Resmini and L. Rosati

and a new longing for community has been aroused. We see that these fuel a thou-

sand successful social networks, and they are successful precisely because they have

become places [55].

5.5.1 Hansel and Gretel or Getting Lost in the Woods

Space plays a special role in the way human beings experience the world, as we

exist primarily in space: It should not come as a surprise that abstract reasoning is

shaped by underlying spatial patterns. This is the case with language, for example

[21], and it is the case with time. The traditional fairy tale offers a little insight in

this relationship.

Tales of spatial discomfort such as the story of Hansel and Gretel [11], where

children are purposely lost in the woods, express a primary tension between place,

the house, the small community, and space, the indistinct dangerous void of the

forest. Having no place in the world means being lost and risking one’s life.

With communities becoming larger, cities boomed and the woods were cut,

tamed, and made into tilled ﬁelds. Technology provided new means for the objec-

tive measuring of time to a new class of time-savvy merchants [25], and discomfort

moved. In the story of Cinderella, it is being late on midnight that loses the protag-

onist, as she effectively runs out of time as the hours pass. Having no place in the

world means being out of time.

In these stories and in common perception t hough, time is still experienced as

eminently spatial and linear: Hypertext is neither [42]. Nonetheless, experimental

research has shown that space is again the primary and most successful cognitive

metaphor used in dealing with digital space [49], and way-ﬁnding and spatial knowl-

edge [52] are strategic tools for navigation and place-making in information space as

well [35]. But since visual cues are even more important in information space as the

constant dynamic evaluation of being on the right track available in physical space

is lacking [26, 32], way-ﬁnding and spatial knowledge have to be supplemented by

berry-picking and information scent.

5.5.2 Berry-Picking

Berry-picking is an information-seeking strategy that is used when performing

active and directed information search [2]. Berry-picking differs from classical

information retrieval models in that it postulates an evolutionary ﬂow and the

use of different seeking strategies at the s ame time, as the user constantly adjusts

and changes

∗∗∗

his strategy as new information is gained or new goals emerge

(Table 5.2).

Berry-picking is an eminently empirical and cross-contextual theory, maintaining

that similitude, metaphor, and metonymy act as the core mechanisms which enable

successful, adjustable information-seeking strategies, and that information-seeking

strategies apply to both physical and digital environments without any signiﬁcant

differences [3].

5 The Semantic Environment 87

Table 5.2 Comparison of berry-picking strategies in primary and digital sources

Berry-picking strategies In primary sources (paper, etc.) In electronic documents

Footnote chasing

(or backward chaining)

People follow up footnotes

found in books and articles

of interest, and therefore

move backward in successive

leaps through reference lists

People follow up the

contextual links inside

the main text of a page or

the related links near it

Citation searching

(or forward chaining)

One begins with a citation,

ﬁnds out who cites it by

looking it up in a citation

index, and thus leaps forward

It is useful to be able to

both browse through the

complete list of citations

and jump to each

resource in detail

Journal run As one identiﬁes a central

journal in an area, he then

locates the run of volumes of

the journal and searches

straight through relevant

volume years

While browsing an

electronic journal,

accessing by date,

subject, author, metadata

is important, as it is the

possibility to access

index and abstracts for

each issue

Area scanning People browse the materials

that are physically collocated

with/near to materials

located earlier in a search

A numerical notation is of

the utmost importance

because of the

correspondence

notation–class–shelf

Subject search

(in bibliographies,

abstracting and indexing

services)

Once the subject under which a

work is classiﬁed is found,

other works in the same area

are added to the

search/browse set

Navigating an electronic

catalog using subjects

(classes and subclasses)

is often not considered

and an underestimated

feature

Author search Author names are used as

landmarks, as searchers use

them to check if any given

author has done further work

on the same topic

This feature is valuable if

connected to the inherent

hyperlinking qualities of

electronic texts and

related to all relevant

metadata (author,

abstract, etc.)

It seems interesting to point out that book browsing is not an innate human

attribute, but rather a learned, skilled, and a recent one as well [3, 7].

5.5.3 Information Scent

The evolutionary search model of berry-picking highlights the importance of

providing users with contextual cues. Since the goals change and there is no straight-

forward path from start to end, users might get lost in the woods. Information scent,

88 A. Resmini and L. Rosati

a part of the larger theory of information foraging [38], suggests that a common pat-

tern of evolutionary biology is to reuse features adapted for one purpose for a second

different purpose when the environment puts different demands on the species. This

is called exaptation [3].

Information scent maintains that users base their searches on contextual indica-

tors pertaining to the local information space to evaluate the usefulness of a resource

or path, in a way human beings and animals are used to when foraging. These

indicators provide the information scent: The stronger, so to speak, the better, as

it probably suggests the availability of larger quantities of food. In the act of brows-

ing, human beings apply a general propensity to sample and select which has been

evolving through millions of years to information space. Foraging behavior might

have simply exapted to browsing and other information-seeking behaviors [3].

5.5.4 Scenario: Sense of Place in the Supermarket

Cyberspace is not a place you go to but rather a layer tightly integrated into the world around

us. (Institute For The Future [19]).

What does favoring berry-picking and information scent in a store actually mean?

Even if undoubtedly logistics and marketing have provided copious literature on

the subject of organizing consumer way-ﬁnding in supermarkets, their stance is ulti-

mately directed toward improving cross-selling and up-selling metrics. Considering

Place implies orienting spaces and goods layouts to favor exploratory behaviors

along different seamless paths to favor the user experience. This opens up the con-

sumer to serendipity [53], effectively exporting to the physical world one of the

most remarkable and signiﬁcant features of digital spaces and the Web (Fig. 5.2).

In the supermarket, berry-picking is enhanced by favoring it through appropri-

ate techniques such as extending the items listed in Table 5.2 to compare against

the s tore itself (Table 5.3) and providing this additional information to consumers

through its semantic layer, either as digital content or proper coloring, labeling, and

signage (Figs. 5.3, 5.4, 5.5 and 5.6). It is important to note that these are possible

search keys to be implemented at the process level (e.g., the supermarket environ-

ment) and not at the object level (e.g., the single product: This does not simply

concern better box labels).

Table 5.3 Berry-picking techniques applied to a supermarket

Berry-picking in documents Berry-picking in the supermarket

Footnote chasing and citation

searching

Search/access for ingredients,

recipes, coupling

Journal run Products of the same line

Area scanning Area/shelf scanning

Subject search Theme paths

Author search Products belonging the same brand

5 The Semantic Environment 89

Fig. 5.3 Luca is a returning user and the supermarket recognizes him. His phone, PDA, or ded-

icated reader shows him his home-compiled or web-compiled shopping list. Image: A. Falcinelli

Fig. 5.4 A map, path, or coordinate system guides him to the different shelves. Image: A.

Falcinelli

90 A. Resmini and L. Rosati

Fig. 5.5 By reading, scanning, or taking a picture of an item, Luca can receive detailed information

about it on his device. Image: A. Falcinelli

Fig. 5.6 When choosing an item, Luca receives suggestions concerning related products, current

sales, other users’ preferences, or his past favorites. Image: A. Falcinelli

5 The Semantic Environment 91

These heuristics improve the information scent as well: If ﬁnding the shelves

themselves is decidedly easier in the physical world, as opposed to what happens

on the Web, for instance, actual navigation is hindered by a longer list of per-

sonal and environmental constraints, including actual motion from one place to

the other or the time at hand, and favoring exploratory behaviors is more difﬁ-

cult. Resilience and Place allow consumers to have a more complete cognitive map

of the supermarket, effectively improving their navigation (see Section 5.6 below

as well).

Information scent in the supermarket is also reinforced by means of invitations:

When consumers have found the products they are looking for, Correlation (see

Section 5.7) is used to suggest other possible choices, Amazon style: “if you like x

maybe you would like y as well,” or “people buying product x also bought product

y” (Figs. 5.5 and 5.6).

5.6 Choice

“I want a pair of jeans–32-28,” I said. “Do you want them slim ﬁt, easy ﬁt, relaxed ﬁt, baggy,

or extra-baggy?” she replied. “Do you want them stonewashed, acidwashed, or distressed?

[...]” I was stunned. A moment or two later I sputtered out something like “I just want

regular jeans [...]” (Schwarz [48]).

The berry-picking and information scent theories postulate the importance of offer-

ing users multiple cues and paths along the way. Current economic theories suggest

that under the inﬂuence of electronic commerce, markets are moving toward a long

tail model: Strong differentiation and personalization, attention to an increasing

number of niches, selling few items to many instead of many items to few [1].

In economics, it is maintained that the set of available alternatives or choices

has to take into account the costs of economic agents as well as the prefer-

ences of the consumers. This may lead to offering more alternatives not always

being better for economic agents than offering fewer alternatives. Search costs

may induce consumers not to search and not to choose when too many alter-

natives are offered, as they mean too many evaluations to reach a satisfactory

answer [22]. Even when a choice is made, an ever-growing number of possibili-

ties can generate a stressful experience: This is what is known as the paradox of

choice [48].

We have a conﬂict: Berry-picking and a long tail economy suggest that users need

more information; information overload and the paradox of choice suggest more

information is actually less information. Information overload is not (only) an issue

of quantity, but rather an issue of quality, as it directly relates to how information

is organized, visualized, and ﬁnally made accessible: Considerations on place and

time, or context are paramount. Ergonomics and Hick’s law [24, 31, 41] offer s ome

help.

92 A. Resmini and L. Rosati

5.6.1 Hick’s Law

Hick’s law provides the mathematical model which explains the paradox of choice

and suggests some countermeasures as well. Given n equally probable choices, the

average reaction time required to choose among them is approximately proportional

to the logarithm to base 2 of the number of choices, plus 1. That is,

Time = a + b log

(n + 1), (5.1)

where a and b depend on context conditions, such as presentation and the user’s

degree of familiarity with the subject. For example, if choices are poorly presented,

both a and b increase, while familiarity only decreases b.

What’s worth noticing is that the correlation between reaction time and possible

choices is expressed logarithmically, that is, it is non-linear. Every time we choose,

we do not consider every available option (linear time), but rather we cluster options

in categories, dismissing progressively a part of them (roughly half of the options

every time). For a and b constant, if the number of options grows, so does reaction

time. Vice versa, given an equal number of choices, a and b inﬂuence reaction time.

This is why necessary caution has to be used for instance when applying Hick’s

law to items in a menu: If a list is unordered, or ordered meaninglessly from the

user’s point of view, no clustering takes place and the user will probably browse

through each item every time. That means that reaction time becomes linear and

the formula loses any utility. An ordered list (for example, alphabetically ordered

lists such as those in a vocabulary or phone directory) on the other hand allow for

such scanning: The user goes to the pertinent letter (e.g., M) and starts reading only

the relevant subset (e.g., all words whose name begins with M). This time the law

applies and reaction time is non-linear. Hence, choosing once from a signiﬁcantly

ordered eight-item menu is quicker than choosing twice from two four-item menus:

Wide structures (less levels) are to be preferred over deep structures (more levels).

Furthermore, Hick’s law clearly shows that reaction time depends not only on the

number of choices but also on the way these choices are presented to users.

5.6.2 Reducing the Load: Organize and Cluster, Focus,

and Magnify

Hence, an organize and cluster principle could be applied to counter the information

and cognitive load brought on by overloaded choices, in two different directions:

By listing menu items using meaningful, self-evident rules, so that users can cluster

items, according to Hick’s law; or by clustering and organizing in levels, since a

Mathematically, with a menu of eight items the law returns: a + b log

8 = a +3b,aslog

8 = 3.

While with two menus each of four items, the law returns: a+blog

4 = 2a +4b,aslog

4 = 2.

For a n exhaustive review of Hick’s law, see [24].

5 The Semantic Environment 93

wide structure offers no advantage over a deep structure when Hick’s law does not

apply.

Another possible strategy relies on contextualization and customization. This is

what the long tail model prescribes: It is not simply about having fewer choices,

as fewer choices as well could lead to dismissal and failed searches [22], but about

showing choices at the right moment in the right way, where right means appropri-

ate, in context [1]. Amazon’s ﬂexible suggestions interface is a good example of this

behavior (who bought x also bought y, if you are interested in z you might be also

interested in w). This strategy successfully relies on the natural human attitude to

sample and select [3] offered by berry-picking: We could describe this procedure as

focus and magnify – ﬁrst you focus a niche, an item, and then you magnify and look

for s imilar items around it (Figs. 5.5 and 5.6).

The results are analogous to those obtained with clustering: Choices are made

easier and simpliﬁed using logical separations.

5.6.3 Scenario: Choice in the Supermarket

The supermarket clusters products and goods.

Everything is initially organized for simplicity: Complexity is available when

needed, as establishing a feeling of simplicity in design requires making complexity

available in some explicit form [27]. Shelves and aisles work as solution clusters

for large task-driven sections such as grocery, kitchenware, or frozen products, but

they stream down to complimentary paths leading to speciﬁc needs such as sale

products, national products, and organic products. This helps to reduce any ini-

tial spatial discomfort and sense of disorientation consumers might experience in

physical space.

The supermarket focuses and magniﬁes products and goods as well using aisle

and shelf labeling, coloring, and layout. Consumers are supported in their personal

goals and not enticed into compulsive shopping behaviors, as these are no longer

required to push sales. Findability and serendipity are s upported and favored as

well, and they are consumer speciﬁc. The shopping experience improves and this in

turn improves brand perception and goodwill, taking away the sense of frustration

and waste of time and money that compulsion on one side and long searches on the

other carry along with them [18].

Clustering and magnifying are digitally enhanced by appropriate information

architectures, for example using a faceted classiﬁcation scheme which allows any

related products belonging to the same facet (in place, in time, and in context) to be

pushed to consumer devices.

5.7 Correlation

You read the Bible? There’s a passage I got memorized. Ezekiel 25:17 [...] I been sayin’

that shit for years. And if you ever heard it, it meant your ass. (Tarantino, Pulp Fiction)

94 A. Resmini and L. Rosati

Quentin Tarantino’s movie Pulp Fiction is a very good example of non-linear story-

telling. The narrative moves in circles and the beginning and the end of the movie

are tied together in what resembles a Moebius strip pattern, with chronologically

contiguous events moved away from one another and then reconnected with a twist

and some minimal, subtle changes [44].

The world depicted accepts a description of space/time radically different from

the one Cinderella lost her shoe in: There is no linearity, but rather circularity, and

the story moves freely back and forth in both time and space. Because of its intrinsic

fragmentation, the movie can also be considered a tightly integrated series of inter-

connected episodes: Citations of other movies are plenty, and everything seems to

be linked.

Pulp Fiction applies the new model of correlations and semantic ties of hyper-

linking to traditional moviemaking and storytelling: Elements belonging to the same

or to different physical and digital collections (movies and pop culture, in this case)

are connected in one single whole, enriching the narrative with supplementary layers

(Fig. 5.1).

5.7.1 Scenario: Correlation in the Supermarket

Implicit recommendations are relationships formed by social behavior – the traces of many

people acting with objects. (Marchionini [30])

In the supermarket all items are correlated, either by agents or by consumers.

Using their cell phones, hand-held devices, or dedicated readers, consumers

receive personalized information concerning theme paths, coupling paths (if you

bought a we suggest b and c that you can ﬁnd on shelf x), recommended or

best-selling products, products on sale.

Since the supermarket is a Place, they receive integrated maps, and GPS-like

itineraries as well, highlighting items of interest and correlated products.

Consumers are an integral part of the human–information interaction process;

they also have access to a wide array of social, collaborative content shared by the

community of people buying in the supermarket. Some consumers have re-cataloged

part of the available products using faceted classiﬁcation, some others provide top

tens, reviews, and link to external informational sources on recipes or health-related

matters, some others have tagged most of the cereals and bakery products according

to their vitamin content. These additional layers are all available and connected to

the larger pervasive information layer: Products acquire histories, annotations, and

links that strongly affect their ﬁnding and use.

Correlations are freely added by all users of the system, allowing opportunities

to discover new kind of features and to relate products and opinions in novel ways

which allow for more productive, synchronic clustering [30].

5 The Semantic Environment 95

5.8 The Semantic Supermarket

In the semantic supermarket, products are classiﬁed by using a mixed classiﬁca-

tion scheme which uses traditional top-down classiﬁcation, faceted classiﬁcation,

and free-form social classiﬁcation accordingly. In a niche or specialized retail store

selling a more homogeneous range of products, it might be possible to use just

one single general faceted scheme, since a pre-requisite for the applicability of

faceted classiﬁcation is that the items to be classiﬁed share similar characteristics

[15, 40]. Facets allow for easy correlation and favor berry-picking. In the semantic

supermarket, this means experimenting with product layouts, for example by hav-

ing rack sections for Meals, Times of the day, or Kitchen, Bathroom and Bedroom,

or Regional recipes.

In the semantic supermarket the general layout might even move away from the

common linear, sequential l ayout where a number of straight racks are placed back

to back, for example to a radial structure where a core, a main hall, acts as the entry

Table 5.4 Continuity between digital and physical environments

Web site Retail shop

General

classiﬁcation

model

Taxonomy

implemented

as navigation

and navigation

aids

Taxonomy implemented as

department, shelves, and

goods layout. Study of

paths

Findability 1

Controlled

vocabularies

Thesauri,

metadata

Product index (with links,

variants, and synonyms),

RFID or SIM enabled

Direct search Search engine

using

vocabularies

Interactive displays, PDAs,

help desks

Findability 2

Shortcuts Quick links,

social

navigation,

personalization

(history, wish

list)

Color-coded paths,

integration with the web

site

Information

circularity

Contextual

navigation