SHOPPER'S EYE:
`
`Using Location-based Filtering for a Shopping Agent in the
`
`
`
`
`
`Physical World
`Andrew E. Fano
`
`Center for Strategic Technology Research
`
`
`
`Andersen Consulting
`3773 Willow Road
`
`Northbrook, Illinois 60062-6212
`
`
`(847)714-2826
`
`afano@cstar.ac.com
`
`shopping goals to Shopper's Eye. Then, as
`
`
`1.ABSTRACT
`
`
`
`
`shoppers stroll through a mall, Shopper's Eye
`Agents of all types rely on easily computed
`
`
`informs them of the availability of items of
`
`
`features that are suggestive of a user's
`
`
`interest to them available in the immediately
`
`
`preferences and goals to define and constrain
`
`
`surrounding stores, as well as any cheaper
`
`
`
`
`their tasks. Although a person's location is
`
`
`local alternatives. Knowledge of both the
`
`
`
`usually suggestive of their current activity,
`
`
`
`user's goals, and the environment in which
`
`
`
`agents have not relied upon a user's location to
`
`
`they act is a powerful combination that
`
`
`constrain their task. This is because users
`
`
`enables Shopper's Eye to bring relevant
`
`
`have typically used their computers only from
`
`
`information to the user throughout the course
`
`
`home or work, and because location has not
`of their task.
`
`
`been easily computable. However the
`
`
`
`explosive growth in the use personal digital
`We begin the paper by introducing the
`
`
`
`
`
`assistants (PDAs), laptop computers, and
`
`
`concept of location-based filtering -exploiting
`
`
`global positioning system (GPS) receivers is
`
`
`
`the user's location to constrain the task of an
`
`
`enabling people to use computers in the most
`
`
`agent. We then discuss how this technique is
`
`remote of locations, and to have their location
`
`
`
`applied in support of physical shopping. We
`
`
`accessed by software.
`
`
`conclude with a discussion of the Shopper's
`
`
`Eye project and a description of the current
`
`Shopper's Eye prototype.
`
`This paper introduces Shopper's Eye, a PDA­
`
`
`
`
`
`based, GPS-enabled agent prototype that
`
`
`
`relies on knowledge of a shopper's physical
`1.1 Keywords
`
`
`
`location to support the shopping task while
`
`
`
`
`
`
`shopping at a mall. Shoppers indicate their
`2.LOCATION-BASED FILTERING
`
`
`
`A central issue for developing agents of all types is
`l'ennission to make digital/hard copies of all or part of this material 1,ir
`
`
`
`
`
`
`
`
`personal or classroom use is granted without kc· prnvitkd that the copies
`
`
`
`
`
`identifying easily computed features that are either very
`
`
`
`
`are not made or distributed for prolit or commercial advantage, the eorv­
`
`
`
`suggestive of the user's preferences and goals or can
`
`
`
`
`
`right notice. the title of the publication and its date appear. and notice is
`
`somehow be used to constrain the task of the agent.
`
`
`
`
`givt!n that copying is hy pennissio11 of J\l'!\1. lnc. ·1·0 copy lllh�rwise.
`
`
`
`
`
`to republish. lo post on servers or t,l redistribute to lists. requires prior
`
`
`
`Keyword-based approaches are commonly used. For
`
`
`
`sp�cilic pennission and/or ke.
`
`
`example, users may be asked to specify keywords to
`Autonomous Agents 98 Minneapolis MN l JSA
`
`
`
`
`
`explicitly identify their goals [6], or keywords and key
`
`
`Copyright I 998 0-89791-983-1 /98/ 5 ... $5.00
`
`
`phrases may be extracted from user data [9]. Collaborative
`
`
`
`
`
`filtering, another technique, involves extending user
`
`Location-based filtering, GPS, shopping, PDAs, agent.
`
`416
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 1 of 6
`
`

`

`rooms in an office, is somewhat similar in its use of
`
`
`specified preferences by incorporating those of other users
`
`
`
`
`
`location-awareness as a means of capturing the user's
`
`
`whose preferences overlap [8]. The demographic
`
`context [10].
`
`
`
`generalization method involves classifying a user using
`
`
`
`minimal user input into demographic categories with well­
`2.1 The Predictive Value of Location
`
`
`
`
`understood preferences [7]. These techniques are all
`
`
`Our physical location is often very predictive of our current
`
`
`
`intended to infer as much as possible about a user's goals
`
`
`task. If we know someone is at a bowling alley or a post
`
`
`and preferences based on observable features, while
`
`
`office we can reasonably infer their current activity.
`
`minimizing the need for user input.
`
`
`
`
`
`Knowledge of a user's current task largely determines the
`This paper introduces Shopper's Eye, an agent that exploits
`
`
`
`
`type of information they are likely to find useful. People
`
`
`
`
`a feature that, to our knowledge, has not yet been applied to
`
`
`
`are unlikely to concern themselves with postal rates while
`
`
`
`
`
`information gathering agents: the physical location of the
`
`
`
`bowling, or optimal bowling ball weight while buying
`
`
`user. SHOPPER'S EYE is an agent running on a PDA
`
`
`stamps. In addition, knowledge of the resources and
`
`digital assistant) equipped with a GPS (global
`(personal
`
`
`
`obstacles present at a particular location suggest the range
`
`
`
`
`positioning system) receiver, intended to support shopping
`
`
`
`
`at that location. of someone of possible and likely actions
`
`
`
`in an outdoor mall. This agent assists shoppers by
`
`
`
`
`
`This awareness of a user's possible and likely actions can
`
`
`
`providing information about merchandise in which they
`
`
`
`be used to further constrain the type of information a user
`
`
`
`
`
`earlier expressed an interest. As a shopper strolls through a
`
`
`is likely to find useful. For example, knowledge of a
`
`
`
`mall, SHOPPER'S EYE alerts him or her to the availability of
`
`restaurant's wine list could be used by a recommender
`
`
`
`
`merchandise of previously specified categories in the
`
`
`
`system to constrain the wine advice it presents.
`
`
`
`
`surrounding stores, as well as any cheaper alternatives in
`Knowledge of a shopper's precise location in a shopping
`
`
`
`
`
`
`
`precise the local area. Shopper's Eye exploits the user's
`
`
`
`
`mall is valuable to Shopper's Eye because it enables the
`
`
`
`
`physical location to filter the information it presents.
`
`
`
`
`identification of the stores immediately surrounding the
`Web agents have not used physical location as a predictive
`
`
`
`
`
`
`shopper. The offerings of the stores closest to the shopper
`
`
`
`feature because the locations from which users access the
`
`
`
`represent the immediate choices available to the shopper.
`
`
`
`
`
`web have largely remained constant - typically their home
`
`Given that shoppers place a premium on examining
`
`
`or office. Moreover, location has not been a particularly
`
`
`merchandise first hand and that there is a cost associated
`
`
`easy feature to compute and unambiguously communicate
`
`
`
`with walking to other stores, the merchandise of the closest
`to an agent.
`
`
`
`
`surrounding stores constitute the most likely immediate
`
`
`
`selections of the shopper. Consequently, among the most
`
`
`
`However, the explosive growth in the use of laptops and
`
`
`useful information Shopper's Eye can provide at any given
`
`
`PDAs signals an important change. As we begin to find
`
`
`time is the availability of merchandise in the surrounding
`
`
`
`
`ourselves bringing our PDAs and laptops everywhere we
`
`
`
`
`
`stores that matches their previously stated goals.
`
`
`go, the particular locations we use them will increasingly
`
`
`
`reflect an important part of our current context.
`We tend to move to different locations while performing
`
`
`
`
`
`
`Furthermore, our precise location can now be passively and
`
`many of our tasks. This suggests our immediate
`
`
`
`
`unambiguously obtained by software through the use of
`
`
`
`surroundings do not completely capture the full range of
`
`
`
`GPS receivers. Such receivers are becoming increasingly
`
`
`options we may have. In fact one of the main reasons for
`
`Some are now available as
`
`affordable and compact.
`
`
`
`leaving a location is to perform an action that is not
`PCMCIA cards.
`
`
`
`possible at our current location. Nevertheless, we do tend
`
`
`
`to address most tasks within relatively local areas. Thus
`Location has, of course, played a significant role in other
`
`
`
`
`
`
`
`
`while our immediate surroundings suggest the options we
`
`
`
`
`areas research. Navigation, most obviously, has relied on
`
`
`view of a have available at a given point in time, a broader
`
`
`
`
`
`the ability to detect and monitor location. Recent work on
`
`
`
`
`location will often capture the options we are likely to
`
`
`
`supporting user mobility in which personalized computing
`
`
`consider over the course of a task. In the case of mall
`
`
`
`
`environments follow users to remote locations also rely on
`
`
`
`
`shopping, for example, the stores immediately surrounding
`
`
`
`knowledge of a user's location (2, 3].
`
`
`
`
`the shopper represent the options available at that moment.
`In these cases, however, location is the problem. That is, a
`
`
`
`
`
`
`
`
`
`Mall shoppers, however, are generally willing to travel to
`
`
`vehicle must be guided from one point to another, or a
`
`
`
`the potential options any store within the mall. Therefore
`
`
`
`computing environment must be replicated at a remote
`
`
`
`
`over the entire shopping trip include all the stores in the
`
`
`
`
`location. In SHOPPER'S EYE the user's location is used in a
`
`
`mall. Accordingly, Shopper's Eye presents offerings of
`
`
`
`
`very different way. Rather than defining the problem, the
`
`
`
`interest only from the immediately surrounding stores
`
`
`
`user's location is a crucial piece of data that can be used to
`
`
`because these are the immediately available options. When
`
`
`
`inform and constrain the information gathering task. The
`
`
`
`
`asked for alternatives Shopper's Eye restricts itself to all
`
`
`
`
`ParcTab based "location browser", which displays file
`
`
`
`the stores within the mall -the area within which the
`
`
`
`directories and runs programs associated with particular
`
`417
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 2 of 6
`
`

`

`A second, related difficulty lies in communicating our
`desires to an agent. Shopping agents are great if we know
`the precise commodity we want. We can simply enter the
`product by name. Unfortunately,
`if we don’t have a
`3. PHYSICAL VS. ONLINE SHOPPING
`specific item in mind when weshop, then the problem of
`It
`is tempting to argue that online shopping will soon
`conveying what we want
`to an agent becomes more
`become the predominant mode of shopping, pending only
`difficult. For example, how doItell an agent what kind of
`greater penetration of home computers, the expansion of
`lamp I want for my living room?
`online offerings, and better online shopping tools. It would
`therefore be a mistake to begin using location to support an
`activity that will become virtualized. Already we’ve seen
`the emergence of a numberof software agents that support
`online shopping. For example BargainFinder
`[6] and
`subsequently Shopbot [4] both allow users to identify the
`cheapest source for a music CD, given a title.
`Similar
`programs have been developed for buying books, such as
`BargainBot [1]. These systems demonstrate the potential
`of electronic commerce web agents to create perfect
`markets for certain products. The success of these agents
`will encourage the development of similar web shopping
`agents for a greater variety of goods.
`
`Undeveloped preferences
`
`to include descriptive
`that allow shoppers
`Interfaces
`features like price ranges, color, options, brands, etc, can
`help address the above problem, but they are not enough.
`Much of
`the
`time
`shoppers
`either haven’t
`formed
`preferences or can’t articulate their desires until after
`they’ve started shopping and had a chance to examine
`various examplesof the target products.
`
`Shopping is entertainment
`
`People like to shop and do so without having a specific
`purchase in mind. One study found that 42% of consumers
`are “non-destination shoppers” that visit the mall primarily
`for leisure browsing and socializing (Kirtland, 1996).
`
`shopping task as a wholeis likely to be performed. Being
`alerted that a store hundreds or thousands of miles away
`sells the same merchandise for a few dollars less than the
`cheapest local alternative is of little value in cases when
`shoppers
`require
`a
`first
`hand
`examination of
`the
`merchandise in question or are not willing to wait for
`shipping.
`
`used, the color, how it fits and feels, and the workmanship.
`Similar problems apply to most other products.
`
`Imprecise goal specification
`
`3.1 The Limitations of Online Shopping
`Certainly online shopping will continue to grow and the
`trend towards more powerful online shopping agents will
`continue. Nevertheless, it also seems clear that no matter
`how sophisticated web-agents become,traditional physical
`shopping will continue to dominate the market for the
`foreseeable future. Several! inherent difficulties of online
`shopping will ensure the continued reliance on physical
`shopping:
`
`Non-fungible goods
`
`Web-based shopping agents have typically enabled users to
`identify the cheapest price for fungible products such as
`books and music CDs. While this capacity to create
`“perfect markets” for such commodities is of great benefit
`to consumers, several difficulties exist that will complicate
`applying these approachesto arbitrary products.
`
`suited to shopping
`Commodities are particularly well
`agents because it
`is easy to make comparisons between
`competing offers. Because commodities are fungible, one
`of the very few dimensions upon whichthey differ is price.
`Price therefore becomes the primary, if not sole, criterion
`upon which purchasing decisions are made.
`
`As soon as we move beyond commodities, however,
`several other criteria become important. For example, how
`do we compare items such as sweaters, mattresses, or
`tables?
`In addition to price we care about the materials
`
`Shopping is sensory
`
`Even if we could effectively provide these details most of
`us would be unlikely to delegate a purchasing decision to
`such an agent. After all, many people are uncomfortable
`even trusting spouses to make appropriate purchases on
`their behalf. Most people want to see and touch first hand
`what
`they’re
`considering before making a purchase
`decision. The few preferences we may provide an agent
`cannot replace this rich, first-hand experience. At best
`such preferences could be used to generate a candidate set
`for shoppers to consider.
`
`Instant Gratification
`
`Shopping is often a very emotional activity. People are
`pleased with their purchases and often can’t wait to get
`hometo try them out. The inherent delay between online
`purchases and their receipt is a significant issue to those
`who simply must take hometheir selections as soon as they
`see them.
`
`In the end, consumers will continue to engage in physical
`shopping because ofthe limitations listed above. However,
`the fact
`that the task can’t completely be delegated to
`software agents does not rule out a role for them. First, of
`course, as the success of programs such as BargainFinder
`have demonstrated, users find them useful for purchasing
`commodities when they know what they want. A second
`role, however,
`is to support
`the physical shopping task
`
`418
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 3 of 6
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 3 of 6
`
`

`

`itself, throughout the time that a person is engaged in it.
`This, of course,
`is the approach taken in the SHOPPER’S
`EYEproject.
`
`4. SHOPPER’S EYE
`At first blush it may seem that SHOPPER’S EYE is subject to
`some of the same limitations as purely web-based agents.
`After all, why should it be any easier to communicate our
`goals to SHOPPER’S EYE than it is to a web-based agent?
`Why would our preferences be any more developed for
`purchases supported by SHOPPER’S EYE than a web-based
`agent?
`
`A key difference between purely web-based agents and
`what werefer to as “physical task support agents” (i.e. an
`agent that supports a user engaged in a task in a physical
`setting) is that web-based agents are completely responsible
`for conveying all information that will be considered by the
`user. On the other hand, “physical task support” agents
`such as SHOPPER’S EYE can augment the approaches of
`web-based agents by referring to aspects of a user’s
`environment. For example, it is not terribly important for
`SHOPPER’S EYE to conveyrichly the feeling of a particular
`sweater if the sweater is in a store thirty feet away.
`It need
`only refer the shopper to the sweater. The shopper will
`gain a much better appreciation of the sweater by trying it
`on than through anything that can be conveyed by the
`system. When too many products match an imprecisely
`specified goal for a web-based agent, a more restrictive
`search must be made.
`In SHOPPER’S EYE, however, many
`matches simply indicatesthere is a store thatis likely to be
`of great interest to the shopper and therefore should be
`visited. Once inside, narrowing down the merchandise of
`interest in person will often be far easier than refining our
`goals on a web-based agent.
`Therefore physical
`task
`support agents like SHOPPER’S EYEcan help users elaborate
`their preferences and identify specific goals by calling
`users’ attention to aspects of their physical environment as
`a means of conveying information throughout the entire
`course ofthe task.
`
`Specification ofgoals
`
`the general
`least
`Shoppers begin by indicating at
`category of merchandise they are interested in. These
`goals may berefined as the task progresses. Shopping
`agents need to enable the specification of goals at
`various degrees of specificity.
`
`Exploration ofProduct Space
`
`Before shoppers can make a selection, they need to
`become educated about what is available. Shopping
`agents can aid in this task by presenting various classes
`of offerings, reviews, demonstrations, etc. Physical
`shopping agents can augment
`this by providing
`shoppers with a tour ofthe locally available offerings.
`
`«
`
`Refinement ofpreferences
`
`As shoppers learn what is available and examine the
`offerings their preferences evolve. Agents need to
`enable shoppersto refine their preferences overtime.
`
`identification and comparison ofcandidate products
`
`As shoppers begin to understand what they want and
`what
`is available they typically compile a list of
`candidates that will be considered more carefully.
`Agents
`should
`support
`the
`construction
`and
`maintenance of such lists and facilitate the comparison
`of candidates within the list according to various
`criteria.
`
`Negotiation ofoffers
`
`Shopping agents need notbe restricted to providing the
`shopper with information.
`Ideally shopping agents
`should negotiate prices and service options with
`retailers.
`
`©
`
`Product Selection and Purchase
`
`shopping agents
`Naturally,
`transaction itself.
`
`should facilitate
`
`the
`
`e
`
`Product Support
`
`The shopping agent should be able to be used as a
`channel
`through which product
`service
`can be
`delivered.
`
`4.1 The Promise of Physical Shopping Agents
`It
`is hardly surprising that physical shopping has been
`neglected by the agents community. After all, until very
`recently there simply was no reliable way to deliver
`is worth emphasizing the potential of such agents to
`It
`customized information to individual shoppers in remote
`operate as bi-directional channels. That is, not only can
`locations. However, the explosive growth of PDAs, and
`they provide information to the shopper, but, at
`the
`
`their—_increasingly sophisticated|communications
`shopper’s discretion,
`they may provide information to
`capabilities promise to make them effective channels of
`retailers as well. We intend to demonstrate this potential in
`“just in time” information to users wherever they happen to
`the next version of SHOPPER’S EYE by having it
`be.
`communicate a shopper’s goals and preferences to a
`retailer-based agent, who,
`in turn,
`responds with a
`customized offer
`that bundles service along with the
`product. Enabling the customization of offers is crucial to
`gaining the cooperation of retailers who are reluctant to
`
`SHOPPER’S EYE is an example of an agent that supports
`physical
`shopping by exploiting the promise of this
`developing channel. We intend SHOPPER’S EYEto support
`all phases of the shopping task including:
`
`419
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 4 of 6
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 4 of 6
`
`

`

`compete solely on price and of value to customers who
`base their purchasesoncriteria other than price.
`
`4.2 The Current SHOPPER’S EYE Prototype
`At
`the present
`time we have a working prototype of
`SHOPPER’S EYE for a Windows CE PDA equipped with a
`GPS(global positioning system) receiver. This prototype
`works for the Old Orchard shopping center
`in Skokie
`Illinois, an outdoor mall containing approximately 110
`stores.
`
`SHOPPER’S EYE uses a GPSreceiver to determine the user’s
`location, thereby limiting its use to outdoor malls. The
`advantage of GPS for SHOPPER’S EYEis that it enables the
`retrieval of data for nearby stores without relying on the
`presence of any special equipment at
`the mall
`itself.
`Although the accuracy of smaller, inexpensive receivers is
`limited to approximately 75-100 feet,
`this has thus far
`proven to beall that is necessary to identify accurately the
`immediately surroundingstores.
`
`to create online catalogs, we can expect the number of
`differing formats to decrease,
`resulting in a tractable
`number of competing formats. As electronic commerce
`progresses,
`it
`is not unreasonable to expect standards to
`evolve
`governing
`how merchandise
`offerings
`are
`represented.
`
`The current version of SHOPPER’S EYEis designed for use
`as follows:
`
`GoalSpecification
`
`shoppers using
`shopping trip,
`leaving on a
`Before
`SHOPPER’S EYEcreate a shopping list of items by selecting
`from a preexisting set of approximately 85 product
`categories (e.g. men’s casual pants, women’s formal shoes,
`flowers, etc,). They also indicate the shopping venue they
`intend to visit from a list of malls. As mentioned above,
`only one mall has been includedatthis time.
`
`SETS ett ee Te bs
`
`Browse Glance below for thems of interest in nearby stores as
`
`Figure 1: The SHOPPER’S EYEscreen.
`
`SHOPPER'S EYE is informing the shopper of items of interest available at Lord and Taylor.
`The presence ofthe “save”icon indicates that a cheaperlocal alternative is available.
`
`The current SHOPPER’S EYE prototype uses generated data
`rather
`than actual
`store
`ads
`and prices. While we
`acknowledge that gaining access to store merchandise and
`prices in a manner that can be compared across varying
`formats is an important
`issue,
`it
`is not the focus of our
`current work. We are optimistic that this challenge will be
`addressed in a numberof ways. First online catalogs tend
`to be reasonably well structured. Already some researchers
`have had success building agents that “learn to shop” at a
`given store using a relatively small amount of knowledge
`[4]. Moreover, as retailers begin to use standard packages
`
`Initial Store Selection
`
`the mall, SHOPPER’S EYE begins by
`Upon arriving at
`suggesting the closest store that sells at least one item of a
`type entered by the user during goal specification. Along
`with the store name SHOPPER’S EYElists the specific items
`available and their prices. A map of the mall displays both
`the precise location of the store and the shopper’s current
`location. The shopper can ask SHOPPER’S EYEto suggest a
`store at any time based ontheir current location.
`
`Browsing
`As noted earlier, 42% of people who visit malls do so
`
`420
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 5 of 6
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 5 of 6
`
`

`

`without a particular destination in mind. As shown in
`Figure 1, SHOPPER’S EYE includes a browse mode for use
`by shoppers as they stroll through the mall.
`In browse
`mode SHOPPER’S EYE suggests items of interest for sale in
`the stores currently closest
`to the shopper. An item is
`considered to be of interest if it matches the categories
`entered in the goals screen. If there are no itemsof interest,
`SHOPPER’S EYE simply states
`the general
`type of
`merchandise sold at that store, rather than specific items.
`As the shopper strolls a map displays his or her precise
`current location in the mall.
`
`Alternatives
`
`If an item displayed is selected by the shopper while
`browsing, SHOPPER’S EYE will alert the shopperto the local
`retailer offering the same product for the lowest price, or
`announce that
`it
`is the best
`local price. This search is
`restricted to the local mall, as that is the assumedradius the
`shopperis willing to travel.
`
`4.3 Future Work
`As mentioned above, we are currently developing a new
`version of SHOPPER’S EYE intended to support the broader
`aspects of the shopping task. Weare particularly interested
`in exploring the possibilities enabled by treating SHOPPER’S
`EYE as a two way channel between the shoppers and
`retailers. While this paper has discussed location-based
`filtering primarily in the context of the shopping task, we
`are confident this technique can help form the basis for
`“physical task support” agents that provide an information
`channel to people engagedin tasks in the physical world.
`
`5. ACKNOWLEDGMENTS
`Thanks to Nagendra Prasad, Kishore Swaminathan and Joe
`McCarthy for their helpful comments and discussions on
`drafts of this paper. Thanks also to Rafael Dowti and Scott
`Kurth for their programmingefforts.
`
`6. REFERENCES
`[1] Aoun, B. Agent Technology in Electronic Commerce
`and
`Information Retrieval
`on
`the
`Internet
`in
`
`Proceedings of AusWeb96 Second Australian World
`Wide Web Conference,
`(Lismore, Australia,
`July
`1996).
`[2] Bacon, J. Bates, J. and Halls, D. Location-Oriented
`Multimedia.
`In JEEE Personal Communications 4, 5.
`(New York, October 1997) 48-57.
`(3] Bharat, K., and Cardellio, L. Migratory Applications.
`In Proceedings ACM Symposium on User Interfaces
`Software and Technology. (Pittsburgh, PA, November
`1995).
`[4] Doorenbos, R. B., Etzioni, O., and Weld, D. S. A
`Scalable Comparison-Shopping Agent for the World-
`Wide Web.
`In Proceedings of the First International
`Conference on Autonomous Agents. (Marina Del Rey,
`Calif. Feb. 1997).
`Kirtland, K. M. The 1996 Mall Customer Shopping
`Patterns Report ICSC Research Quarterly, 3, 4. 1996.
`{6] Krulwich, B. The BargainFinder agent: Comparison
`price shopping on the internet.
`In Williams, J., ed.,
`Bots and Other Internet Beasties. SAMS.NET, 1996.
`
`[5—
`
`Intelligent User
`LIFESTYLE FINDER:
`[7] Krulwich, B.
`Profiling Using Large-Scale Demographic Data.
`In Al
`Magazine. 18, 2. (1997) 37-45.
`{8] Lashkari, Y., Metral, M., & Maes, P. Collaborative
`Interface Agents.
`In Proceedings of the Twelfth
`National Conference on Artificial Intelligence (Menlo
`Park, Ca. 1994) 444-449.
`[9] Pazzani, M., Murumatsu, J. and Billsus, D.. SYSKILL
`& WEBERT:
`Identifying Interesting Web Sites.
`In
`Proceedings of the Thirteenth National Conference on
`Artificial Intelligence (Menlo Park, Ca. 1996) 54-61.
`{f0]Schilit, B. N., Adams, N. and Want, R. Context-
`Aware Computing Applications. In Proceedings IEEE
`Workshop
`on Mobile Computing Systems
`and
`Applications. (Santa Cruz, CA, December 1994) 85-
`90.
`
`421
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 6 of 6
`
`Google, Exhibit 1014
`IPR2022-00742
`Page 6 of 6
`
`