This paper discusses the concept of information market. The authors of this paper have been involved in several aspects of information retrieval research. In continuing this research tradition we now take a wider perspective on this field, and position it as a market where demand for information meets supply for information. Our modern day western societies are dominated by information systems. This is not a new phenomenon, as there were information systems playing important roles in cultures and empires long since gone. The problem of managing large volumes of information is also not new. The rfist institutional libraries appeared in Athens during the 4th century BC. Around that time in the library of Alexandria the rfist catalogs were used. The Romans later introduced classification and in the 18th century Dewey elaborated on that with the Dewey Decimal System. These days we use computers to assist us in managing these large volumes of information, either stored in a physical bricks and mortar library or on the Web. Hand in hand with the increased amounts of information that needed processing, the problem of information overload started to surface. As more and more data accumulated in information systems, it became harder and harder to nfid those bits of data that really mattered. This has led to the introduction of the field of information retrieval [1]. The development of the Internet provided our society with the opportunity to interconnect computers, leading to networked information systems. As the Internet matured it gave birth to the World-WideWeb (the Web). This resulted in a multiplication of the information available to people around the globe and gave birth to e-commerce. Given the abundance of information available via the Web, an important part of the commodities traded on the Internet are actually carriers of information. This paper proposes to look at the exchange of information on the Internet as an Information Market, where demand and supply of information meet. As such, our aim is to mark a transition from a traditional view on information retrieval to an Information Market Paradigm. A traditional perspective on information retrieval is provided in the Information Retrieval Paradigm [2]. On one side, there are information resources that are at our disposal. These resources, which may be aggregated,
are characterized in some way to facilitate their discovery. Facing the information carriers is the user with an information need. The user expresses this need in terms of an information request; a query. The query will usually only be a crude description of the actual carrier(s) needed to fulfill the given information need. The need for information will most often be due to some gap in the user’s knowledge. Relevant information is discovered and then absorbed by the user to fill the knowledge gap. In our research we are not concerned with developing yet another approach or strategy to match the demand and supply of information, but rather with an attempt at fundamentally understanding the workings of the Information Market. 2
Our generalized perspective on markets as presented here, is partially based on
the concept of economic markets, in particular on the efild of micro-economics.
We consider economic markets to be a specicfi class of markets dealing with the
trading of goods, services and money. Our considerations are indeed inspired by
literature on economic theories primarily based on the work reported in [
In our view, two main classes of assets can be traded on a market. Ownership of entities, such as physical goods, bank notes, part of an organization, land, etc. The second class, execution of services, pertains to services that may be applied on/to/over entities that are regarded by some participant as value adding, for example treatment of an illness, management of a stock portfolio, etc. In markets dealing with trading of physical goods (i.e., entities) we take the view that what is actually traded is the ownership of these entities. The class of executable services could be split further into Transformation of entities and Reduction of uncertainty. Let us now, however, first explore markets in more detail. We will do so by discussing four core concepts: transactions, cost/benefits, preference and value addition. 2.2
If p1 and p2 are two participants of a market that decide to trade two assets, a1 and a2. This trading of assets is a transaction. In economic markets, transaction participants are said to be either a selling or buying participant. In our view, the notion of selling and buying can only be denfied relative to a specific asset that is involved in the transaction. The sales of an asset by one participant to another participant, will be referred to as a transactand. Let t be a transactand, then we will use t : s →−a b to denote the fact that in transactand t participant s sells asset a to participant b. The (two) participants in a transactand are given by the function Participants(t) = {s, b}. Similarly, the buyer and seller ‘role’ within a transactand are given by Buyer(t) = b and Seller(t) = s respectively. A transaction can now be regarded as being a set of transactands. If T is a transaction, then we can define: a a bi .
b ∈ T , ∃t∈T ht : s →− s →− As a rule we will require: t1, t2 ∈ T ∧t1 : s →−a b ∧ t2 : s →−a b ⇒ t1 = t2. In other words, the involved participants and asset uniquely determine the transactand in a transaction. This will allow us to denote the initial transaction between p1 and p2 as: {p1 →−a1 p2, p2 →−a2 p1} and a more complex set of transactions involving p3, p4 and p5 as: {p3 →−a3 p5, p5 →−a4 p4, p4 →−a5 p3}. The set of participants involved in a transaction are defined as: Participants(T ) , ∪t∈T Participants(t). This leads us to the question why do the transactions take place in the first place? There is usually some benefit to the participants of a transaction, therefore a transaction is not just any set of transactands. Each participant in a transaction must both receive and pay an asset:
p ∈ Participants(T ) ⇒ ∃t1,t2∈T [Seller(t1) = p ∧ Buyer(t2) = p] Also, transactions are assumed to be ‘singular’ in the sense that participants of a transaction play the buyer and seller role exactly once. Even more so, a participant can not play the buyer and seller role in one transactand: t ∈ T ⇒ Seller(t) 6= Buyer(t).
We presume the participants of the market to behave in a goal-driven manner. These goals might be explicit in the reasoning of the participants, but may also be more implicit and based on emotions. For the moment we presume GL to be the set of possible goals. Let furthermore, PA be the set of participants on the market and ST be the set of states a participant may hold. A state, in this context, is denfied to be the present satisfaction (of a searcher) with regard to the goals in GL. We presume the function: Id : ST → PA to identify which states belong to which participant. Given the state s of a participant Id(s), we can view the satisfaction of the goals which the participant (in a certain state!) may have as a function: Satisfaction : ST × GL →[0.. 1]. For each goal, the level of satisfaction is expressed as a number between 0 and 1. The consumption of some asset by a participant in a transaction, will result in a change of state of that participant. If T is a transaction, and s is a participant state, then s n T is the state which results after the participation of Id(s) in transaction T . We require the resulting state to belong to the original participant: Id(s) = Id(s n T ) and the participant to indeed be a participant of the transaction: Id(s) ∈ Participants(T ). On closer consideration, our statement: p1 →−a p2 as an abbreviation for: “Participant p1 sells asset a to participant p2” is not specific enough. An actual transaction will take place between participants who hold a specific state. For our considerations in the next subsections, we will need this more refined view. We will therefore use t : s1 →−a s2 as an abbreviation for: “In transactand t, participant Id(s1) in state s1 sells asset a to participant Id(s2) in state s2”. We do require: t : s1 →−a s2 ⇒ t : Id(s1) →−a Id(s1) such that the set of states involved in a transaction is identified as:
States(T ) , ns1 ∃s2,a hs1 →−a s2 ∈ T ∨ s2 →−a s1 ∈ T i o
The actual benetfi of an asset is difficult to measure and defining a measurement
for a certain type of phenomenon is often difficult. Consider the following
historical example as described in [
Benefit(s, T ) , λ g∈GL.max(Satisfaction(s n T, g) − Satisfaction(s, g), 0) We have employed the Lambda calculus notation to denote a function ranging over GL. The costs of an involvement in a transaction can be denfied as the negative impact on the satisfaction levels of a participant:
Cost(s, T ) , λ g∈GL.max(Satisfaction(s, g) − Satisfaction(s n T, g), 0) Given a relative prioritization of the different goals, a weighed level of satisfaction could be computed. Let Priority : ST × GL →[0..1] therefore be a function which identified the level of priority a participant (in a specific state) gives to the speciefid goal. We presume the priority function to be a distribution totaling to one for each of the states: ∀s∈ST [Σ g∈GL Priority(s, g) = 1]. With this weighing function, we can denfie the overall satisfaction as follows:
Satisfaction(s) , Σ g∈GL Satisfaction(s, g) × Priority(s, g) It’s sensible to presume that the level of satisfaction of all participants of a transaction should not decrease: ∀s∈States(T ) [Satisfaction(s) ≤ Satisfaction(s n T )] 3
This section is concerned with a specialization of the ideas presented in the previous section to the context of the information market. It will also position some of the pre-existing research relative to the notion of the information market. 3.1
In accordance to [
Transactions on the information market as such, will not differ dramatically
from markets in general. However, in the case of the information market, we can
elaborate more on the goals which drive the consumers of information resources.
A future consumer of an information resource will have a need for information.
This need for information can be caused by a number of reasons. At the moment
we distinguish between two types of goals: increment of knowledge and change
of mood. The former corresponds to a situation where someone nfids that they
are lacking some information/knowledge. This knowledge gap [
The costs and benetfis of an information resource are particularly difficult to measure. We shall adopt a multi-dimensional view on measuring the potential benetfi of a resource; Utility - dealing with the information that may be provided by a resources and the timeliness, Structure - concerned with the form (report, painting, movie, audio) and format (PDF, MP3) of a resource, Emotion - dealing with the emotional effect (pretty/inspiring) that a resource may have when it is consumed. 4
At the start of this paper we have discussed how an evolution can be observed moving beyond the traditional information retrieval paradigm to an information market paradigm. We have provided a discussion on the general notion of a market where assets are traded. This was then specialized to information resources, leading to an information market. At present, we are working on a more fundamental understanding of markets in general and information markets in particular. Based on these insights, we will evolve our existing theories for different aspects of information retrieval. We expect that models for goal-driven reasoning of participants in the information market will in particular be fruitfull in improving the workings of the information market. Most importantly, we expect this to be most helpful in the retrieval of relevant information by searchers in the information market.