REQUIREMENTS FOR PERSONAL KNOWLEDGE MANAGEMENT

TOOLS

Max V

olkel and Andreas Abecker

FZI Forschungszentrum Informatik Karlsruhe, Haid-und-Neu-Str. 10-14, Karlsruhe, Germany

Keywords:

Personal Knowledge Management, Requirements, Literature.

Abstract:

Personal knowledge management (PKM) is a crucial element as well as complement of enterprise knowledge

management (EKM) which has been largely neglected by Enterprise Information Systems, up to now. This

paper collects requirements for a speciﬁc class of PKM software, which supports personal note taking and

the idea of extending the human memory by information management. It introduces the knowledge-cue life

cycle which describes how information artefacts can be used for helping to denote, remember, use, and further

develop knowledge embodied in people’s heads. Based on this life cycle and on a literature study, this paper

derives a comprehensive requirements catalogue to be fulﬁlled by knowledge articulation tools used in PKM.

This requirements list can be used as a design speciﬁcation and research agenda for PKM tool builders, and to

assess the suitability of existing tools for PKM.

1 INTRODUCTION

As our globalized economy becomes increasingly

more knowledge-based, Enterprise Knowledge Man-

agement (KM, c.f. Probst et al., 2006), (Nonaka and

Takeuchi, 1995), (Abecker, 2004) has become a set-

tled management discipline in the recent 15 years.

The most important area of knowledge creation

and processing is still the most underdeveloped area

of KM: namely that of the individual knowledge

workers’ personal knowledge.

Practically, the main goal of Personal Knowledge

Management (PKM, c.f. Davenport, 2005), is to make

the individual more productive - and thereby also the

organisation as a whole. PKM investigates the use of

methods and tools to amplify the abilities of the indi-

vidual to work better with knowledge (V

olkel, 2010).

Typical use-cases in PKM are, e.g., note-taking, doc-

ument creation, argumentation, idea management, or

managing the personal social network. Many PKM

tools have been built and even more tools are (ab-

)used for PKM tasks. Hence the question arises how

an ideal PKM system look like.

This paper considers the sub-class of PKM sup-

port systems which deal with articulating (parts of)

the individual knowledge in the form of electronic

artefacts.

2 PROCESSES IN PKM

Creating (semi-)formal knowledge is the act of mod-

elling. This paper uses the term “personal knowl-

edge model” to denote an artefact which represents

a set of “knowledge cues”. The knowledge cues can

vary in size, structuredness and degree of formality.

A knowledge cue is either (a) a piece of content,

containing plain text, semi-structured text, or arbi-

trary binary content such as images or desktop ob-

jects, or (b) a connection between other knowledge

cues. Such connections can be unspeciﬁed relations,

directed hyperlinks and formal statements.

This section introduces a novel process model for

the management of knowledge cues. In PKM, orga-

nizer and retriever is the same person. Different from

an organisational context, there is a personal motiva-

tion to organize knowledge. The user can freely trade

efforts of authoring with efforts of retrieval.

An in-depth analysis of economic factors in PKM

has been published by (V

olkel and Abecker, 2008).

Building on this work, a novel PKM process model,

the knowledge cue life-cycle, has been created (c. f.

Fig. 1). It describes ways in which an individual in-

teracts with his personal knowledge model. It consists

of ten processes:

1. Create knowledge mentally;

332

Völkel M. and Abecker A. (2010).

REQUIREMENTS FOR PERSONAL KNOWLEDGE MANAGEMENT TOOLS.

In Proceedings of the 12th International Conference on Enterprise Information Systems - Information Systems Analysis and Speciﬁcation, pages

332-337

DOI: 10.5220/0002969303320337

 SciTePress

reflect

KnowledgeCueLife‐Cycle– collaboratives/w

createknowledge

augment

(content,

structure,and

formality)

codify

retrieve

use

export,

receive,

import

knowledgemodel

Figure 1: Knowledge Cue Life-Cycle.

2. Codify by creating initial knowledge cues;

3. Augment knowledge cues by adding more content,

structure, or formality;

4. Export knowledge cues into other formats;

5. Share knowledge cues with others;

6. Retrieve

knowledge cues and evoke a previously

experienced mind-state;

7. Receive knowledge cues into other formats and

evaluate and ﬁlter them;

8. Import relevant and reviewed parts as knowledge

cues;

9. Use knowledge in a real world situation;

10. Reﬂect on all knowledge cue processes.

3 REQUIREMENTS

This section lists requirements gathered from a vari-

ety of sources analysing PKM.

Of course, no such list can be considered “com-

plete” – there might be even more requirements.

• R1: Knowledge Model should be a Super-set of

existing Conceptual Models. To re-use content re-

siding in one kind of representation in another tool,

This process can also search other people’s knowledge

models, if their knowledge cues are shared.

For this paper, many sources have been used in a non-

systematic manner (papers emailed from colleagues, fol-

lowed forward- and backward references, search engines,

...). The following additional sources have been used sys-

tematically: (a)Proceedings of the Annual ACM Confer-

ence on Research and Development in Information Re-

trieval from the years 2005, 2006, and 2007. (b) Proceed-

ings of SIGCHI. All years up to 2008 have been queried

for “personal”, “knowledge”, and “management”. (c) Pro-

ceedings of workshops on personal information manage-

ment from 2006 until 2009.

Due to very restricted space, a longer version of this

paper is available from http://pubs.xam.de

it needs to be transformed. Transformations between

data-models come not for free. A naive approach to

convert between n formalisms would require writing

transformations. However, if a common interme-

diate formalism can be used, the costs come down to

2n. Learning a new tool requires a user to understand

the conceptual model of the tool. It is easier to learn

a tool with a familiar conceptual model. Hence, the

formalism of a good PKM tool should be similar to

those of other tools in use.

A knowledge model should be expressive enough

to represent existing application data models to enable

re-use of existing external structures. To save costs

in content transformation, the conceptual knowledge

model should therefore be a super-set of the concep-

tual models of all other relevant PKM tools.

• R2: System should Run Queries Automatically.

req:autoqueryauto-query The best personal knowl-

edge model is of no value if a user needs some knowl-

edge and simply forgets to search his personal knowl-

edge model. Therefore, a PKM system should run

queries automatically (R2). (Cutrell et al., 2006) pro-

poses to automatically start a search when certain trig-

gers are encountered. Knowledge cues relevant to the

current task context should be delivered pro-actively

(Schmidt, 2009), relevant to the current business pro-

cess (Abecker et al., 2001).

• R3: Fast Entry. In order to minimize distractions

by one’s own creative new ideas or external inter-

ruptions (phone calls, instant messenger, or colleague

coming in), one needs to quickly take a note and con-

tinue working at the previous task.

• R4: Informal Articulation. Users needs a simple

way to express content in an informal way, e.g., as

plain text, formatted text or box-and-arrow diagrams

(Oren, 2006) or “this is nested within that, but I can’t

say why”.

(Blandford and Green, 2001) studied the use of

short personal notes for task work and found that

informal tools like paper and unstructured text ﬁles

REQUIREMENTS FOR PERSONAL KNOWLEDGE MANAGEMENT TOOLS

333

were sometimes preferred over traditional PIM appli-

cations because they supported more freeform input.

• R5: Formal Articulation. Formal reasoning can

help to reduce retrieval costs, when from a set of ex-

plicitly stated formal statements further formal state-

ments can be inferred automatically.

• R6: Let the User decide on Granularity

of Modelling. req:granularitygranularity Content

varies greatly in size and type. Polanyi (Polanyi,

1998): “... linguistic symbols must be of reasonable

size, or more generally that they must consist of easily

manageable objects. [...] Language can assist thought

only to the extent to which its symbols can be repro-

duced, stored up, transported, re-arranged, and thus

more easily pondered, than the things which they de-

note.”

(Shneiderman, 1989) reports on a comparative

study in which two groups of people had to locate an-

swers to a series of questions in a Hyperties database.

The group with more (46 instead of 5), shorter (4-83

lines instead of 104-150 lines) articles answered more

questions correctly and took less time to answer the

questions.

• R7: Entities need to be Addressable. To be able

to link two entities, they must be addressable. To be

able to model different versions of an entity, each ver-

sion needs a kind of address. In a personal knowledge

model, each entity that is shown to the user should be

addressable.

• R8: Simultaneous Use of Different Levels of

Formality. People need to be able to work at any

level of formality (unstructured, structured and for-

mal knowledge), and to freely mix such levels (Leth-

bridge, 1991). For textual content this means exploit-

ing syntax, structures and semantics. E.g., in semantic

wikis (cf. (V

olkel and Schaffert, 2006)) all three lev-

els can be used. While typing text, there is syntax for

formatting (bold, italic), structures (headlines, lists)

and semantics (typed links).

• R9: Stepwise Changes from Informal to more

Formal Structures. The user should be able to mi-

grate the knowledge into more formal structures, if

desired (cf. (V

olkel and Haller, 2006)). A very im-

portant characteristic of formal knowledge engineer-

ing in general is the modelling process. During mod-

elling, a knowledge model might be in an inconsistent

state. A tool should not simply prevent such inconsis-

tent states, but rather inform the user about the con-

sequences. The migration from one consistent formal

state to another consistent formal state can be a com-

plex operation which cannot or should not need to be

carried out completely in the mind of the user. In-

stead, a sequence of modelling operations should al-

low producing inconsistent states, and then, step-by-

step, move towards the desired target state. Inconsis-

tent states should also be share-able, so that a resolu-

tion can be found collaboratively, if desired.

(Oren, 2006) advises to focus on simply captur-

ing and representing the things that the user wants to

store, before doing any reasoning with it.

•t R10: Knowledge Model Refactoring. With

stepwise formalization a user can gradually add more

structure and formality to knowledge cues. Exter-

nalised personal knowledge artefacts are usually or-

ganised in a systematic manner, e.g., ﬁles are sorted

in folders and sub-folders.

Unfortunately, a good structure today is not a good

structure tomorrow, therefore personal organisation

schemes change. Stuart K. Card in (Jones et al., 2006)

sees this not only as a tedious maintenance task, but

says “re-representation of information is a key to in-

terpreting it”. (Schreiber and Harbo, 2004) empha-

sises ﬂexibility of knowledge models and the need for

reorganisation.

• R11: Versioning. req:versioningversioning The

cost of creating and manipulating knowledge cues is

lower, if people have an easy way to undo their oper-

ations and revert to previous versions of a knowledge

model.

• R12: Capture the Context for Knowledge Cue

Creation and Import req:contextcontext Under-

stand the notion of context, capture it together with

the information and use it to enhance recall and un-

derstanding (Oren, 2006). Automatic context tracking

should relive the user from maintaining bookkeeping

data such as creation data of items or linking (R20)

two items that are commonly used together (Grac¸a Pi-

mentel et al., 2000).

• R13: Active Assistance in Maintenance Tasks.

Metadata about the usage of the knowledge cues by

the system is required: How often did the knowledge

cue appear in search results? How often has it been

changed? When has the most recent statement been

made about this knowledge cue? When was the last

time this knowledge cue was used for inferencing?

Such metadata can be used by the system to ask a

user speciﬁcally and actively about the status of cer-

tain knowledge cues.

• R14: Easy to Learn. Each new tool has a learn-

ing curve that depends on the complexity of the un-

derlying concepts and the user interface. A good user

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

334

interface cannot compensate for an ill-designed un-

derlying data model.

• R15: Grouping of Items. Users need composi-

tion for navigation (Frank, 1988). This allows, e.g.,

browsing and thereby narrowing down their view and

allows discovering related, yet unexpected items.

It is important for a user to be able to group seem-

ingly unrelated content together, so that retrieval of

one item triggers retrieving of the others, too (Jones

et al., 2005).

Grouping knowledge cues is also a pre-requisite

for any kind of batch operations such as exporting,

sharing, refactoring, delete and copy operations.

• R16: Containment Relationship. A containment

relationship is a stronger form of grouping with ad-

ditional semantics for operations. Delete and copy

commands on containers trigger recursively the same

command for all contained elements.

• R17: Optional Naming of Knowledge Cues.

The data model should allow giving things human-

usable names. A name is understood as a unique

name. Names make linking much easier, as the link

target name can simply be typed and one has not to

select from a complex GUI. Names also allow direct

navigation deep into a knowledge model.

But as users often have difﬁculties to ﬁnd names

(Boardman, 2004)[p. 105], (Frank, 1988) advises to

not require a user to name all items.

• R18: Alternative Names. Many systems with

unique names have also means to create additional

alias names, which redirect to another unique name.

• R19: Order Knowledge Cues Ordering a col-

lection of ideas or text snippets into a coherent ﬂow

is one of the main tasks of authoring (Esselborn-

Krumbiegel, 2002). A user should be able to create

order gradually and partially. Note how different this

is from providing a list data-structure: A list can only

represent a total order.

• R20: Linking. (Oren, 2006) ﬁnds “an under-

utilization of the interlinked nature of the informa-

tion”. Knowledge models should allow for precise

and effective linking – and browsing (R27.

• R21: Hierarchy. (Shneiderman, 1996) empha-

sizes the need to get “Overview ﬁrst, zoom and ﬁlter,

then details-on-demand.”

Hierarchies of all kind are commonly used in user

interfaces to let the user narrow down his interests

step-by-step.

• R22: Simultaneous Use of Multiple Levels of De-

tail. Users need ways to see multiple levels of detail

at once (Frank, 1988).

• R23: Annotating Content. When using docu-

ments, a ﬁeld study of (O’Hara and Sellen, 1997) con-

cludes that annotating documents is frequently a part

of the document reading and understanding process.

• R24: Tagging. Tagging is the basic assignment of

easy-to-type keywords to information artefacts. Tag

names contain usually no whitespace and tend to be

really short. A common representation is a tag cloud,

showing all tag names at once, with a font size pro-

portional to their usage frequency. People have prob-

lems in using strict hierarchies (Oren, 2006). There-

fore less strict methods such as tagging (R24) and cat-

egories (R25) are required.

• R25: Classifying into Categories. Categories

differ from tags: Categories tend to have longer,

encyclopedia-like names. In most category-systems,

there is a weak hierarchy, i.e., categories can often

be nested into (several) other categories. In prac-

tice, the boundaries between tags (short, easy-to-type,

not nested) and categories (nice to read, nested) are

blurred sometimes.

• R26: Queries. Besides browsing a user also

needs the ability to search and query the data (Frank,

1988). A number of different queries are re-

quired (fulltext queries, structured queries, aggregat-

ing queries, metadata queries, and formal queries).

• R27: Following Links and Browsing Collec-

tions. Following links is one of the three core strate-

gies of information retrieval described by (Bates,

2002). There is a fundamental difference between

search (where you know what you are looking for)

and browsing (where you ﬁnd things that you placed

there) (Jones et al., 2005).

• R28: Inverse Relations. Many wikis allow

traversing hyperlinks not only forward, but also in

backward mode. For each page they list all pages

linking to the page. In most semantic GUIs incom-

ing links are rendered different from outgoing links.

Therefore it makes a difference for browsing whether

a user stated ([SAP], [employs], [Dirk]) or ([Dirk],

[works for], [SAP]). For the user, this is often an ar-

tiﬁcial distinction. It is desirable that link types have

labels for both directions, e.g., “works for” and “em-

ploys”.

REQUIREMENTS FOR PERSONAL KNOWLEDGE MANAGEMENT TOOLS

335

• R29: Flexible Schema. The survey paper

of (Oren, 2006) states a requirement for ﬂexible

schemas: Leave users their freedom and do not con-

strain them into rigid schemas. This is also relevant

for importing from other data models to be able to rep-

resent as much of the given structure and formality as

possible.

• R30: Transclusion. Embedding a reference and

rendering the content is called transclusion. The

need for transclusion is further explained by (Lud-

wig, 2005), (Nelson, 1995) and in the evaluation of

Popcorn described by (Davies et al., 2006).

• R31: Meta-modelling. If knowledge cues be-

come old, but not outdated, they become just harder

to understand. The meaning of terms shifts. It is

therefore required to let the user describe and anno-

tate all aspects of knowledge cues. Even annotation

on annotations, statements and relations are sometime

required. This allows a user to create a more self-

describing knowledge model. The data model must

allow annotating (and therefore addressing) all of its

elements, in order not to limit expressivity.

4 CONCLUSIONS

This paper investigated requirements for Personal

Knowledge Management tools. A knowledge-cue life

cycle was introduced, which describes how knowl-

edge workers use tools to create information artefacts

that help them to express, remind, share, discuss, use,

and further develop their personal knowledge. Based

on the ten processes, an exhaustive requirements list

has been compiled from existing literature.

Some conclusions and observations can be drawn

from this list of requirements: (1) Ultimately, a

PKM tool must be a general purpose modelling tool:

It must have a rather generic (i.e., not restricted to

a particular domain, R29) data model (R1) with dif-

ferent levels of formality (R4, R5, R8) and granular-

ity (R6). It should allow a user to model a number

of conceptual constructs, namely order (R19), hyper-

links (R20), hierarchy (R21), and annotations (R23).

(2) Even more advanced modelling features such

as inverse relations (R28) and meta-modelling (R31)

are desirable. (3) Some requirements 2, 10, 13,

and 22 can only be realised in a tool, i.e., a model by

itself cannot ’run queries automatically’ (R2), only a

tool can actually run something actively.

The derived list of requirements can be used:

(1) to guide the design of future PKM systems as

well as the underlying research; as well as for

(2) evaluating the adequateness of existing tools

for PKM.

Along the requirements described in this paper, an

initial web-based prototype system has been devel-

oped and evaluated. It is described in an upcoming

dissertation by (V

olkel, 2010). A desktop-based sys-

tem, iMapping, tackling the same requirements is cur-

rently being developed by (Haller, 2006).

REFERENCES

Abecker, A. (2004). Business-Process Oriented Knowledge

Management: Concepts, Methods and Tools. PhD

thesis, Universit

at Karlsruhe (TH), Institut AIFB, D-

76131 Karlsruhe.

Abecker, A., Bernardi, A., Ntioudis, S., Mentzas, G., Hert-

erich, R., Houy, C., M

uller, S., and Legal, M. (2001).

The decor toolbox for workﬂow-embedded organiza-

tional memory access. In ICEIS (1), pages 225–232.

Bates, M. (2002). Speculations on browsing, directed

searching, and linking in relation to the bradford dis-

tribution. In Emerging frameworks and methods: Pro-

ceedings of the Fourth International Conference on

Conceptions of Library and Information Science (Co-

LIS 4), pages 137–150, Greenwood Village, CO. Li-

braries Unlimited.

Blandford, A. E. and Green, T. R. G. (2001). Group and

individual time management tools: What you get is

not what you need. Personal Ubiquitous Comput.,

5(4):213–230.

Boardman, R. (2004). Improving Tool Support for Personal

Information Management. PhD thesis, Imperial Col-

lege, London.

Cutrell, E., Dumais, S. T., and Teevan, J. (2006). Searching

to eliminate personal information management. Com-

munications of the ACM, 49(1):58–64.

Davies, S., Allen, S., Raphaelson, J., Meng, E., Engleman,

J., King, R., and Lewis, C. (2006). Popcorn: the per-

sonal knowledge base. In Conference on Designing

Interactive Systems, pages 150–159.

Esselborn-Krumbiegel, H. (2002). Von der Idee zum Text.

Eine Anleitung zum wissenschaftlichen Schreiben.

Utb. 2. Auﬂage.

Frank, G. H. (1988). Reﬂections on notecards: seven is-

sues for the next generation of hypermedia systems.

Communications of the ACM, 31(7):836–852.

Grac¸a Pimentel, M. d., Abowd, G. D., and Ishiguro, Y.

(2000). Linking by interacting: a paradigm for au-

thoring hypertext. In HYPERTEXT ’00: Proceedings

of the eleventh ACM on Hypertext and hypermedia,

pages 39–48, New York, NY, USA. ACM Press.

Haller, H. (2006). iMapping – a graphical approach to semi-

structured knowledge modelling. In Rutledge, L., edi-

tor, Proceedings of the The 3rd International Semantic

Web User Interaction Workshop (SWUI2006). Poster

presented at the The 3rd International Semantic Web

User Interaction Workshop.

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

336

Jones, W., Phuwanartnurak, A. J., Gill, R., and Bruce, H.

(2005). Don’t take my folders away!: organizing per-

sonal information to get things done. In CHI Extended

Abstracts, pages 1505–1508.

Jones, W., Pirolli, P., Card, S. K., Fidel, R., Gershon, N. D.,

Morville, P., Nardi, B. A., and Russell, D. M. (2006).

”it’s about the information stupid!”: why we need a

separate ﬁeld of human-information interaction. In

CHI Extended Abstracts, pages 65–68.

Lethbridge, T. C. (1991). A model for informality in knowl-

edge representation and acquisition (an extended ab-

stract). In Workshop on Informal Computing, Santa

Cruz, Incremental Systems, pages 175–177.

Ludwig, L. (2005). Semantic personal knowledge manage-

ment. Technical Report D11.01 v0.01, DERI Galway,

University Road, Galway, Ireland.

Nelson, T. H. (1995). The heart of connection: hypermedia

uniﬁed by transclusion. Communications of the ACM,

38(8):31–33.

Nonaka, I. and Takeuchi, H. (1995). The Knowledge-

Creating Company : How Japanese Companies Cre-

ate the Dynamics of Innovation. Oxford University

Press.

O’Hara, K. and Sellen, A. (1997). A comparison of reading

paper and on-line documents. In CHI, pages 335–342.

Oren, E. (2006). An overview of information management

and knowledge work studies: Lessons for the seman-

tic desktop. In Decker, S., Park, J., Sauermann, L.,

Auer, S., and Handschuh, S., editors, Proc. of the Se-

mantic Desktop and Social Semantic Collaboration

Workshop, volume 202 of CEUR Workshop Proceed-

ings ISSN 1613-0073. CEUR-ws.org.

Polanyi, M. (1998). Personal Knowledge. Routledge.

Probst, G., Raub, S., and Romhardt, K. (2006). Wis-

sen Managen: Wie Unternehmen ihre wertvollste

Ressource optimal nutzen. Gabler Verlag, 5 edition.

Schmidt, A. (2009). Situationsbewusste Informations-

dienste f

ur das arbeitsbegleitende Lernen. Karl-

sruhe Institute of Technology / Universit

at Karlsruhe

(TH), Karlsruhe, Germany. http://digbib.ubka.uni-

karlsruhe.de/volltexte/1000012939.

Schreiber, T. and Harbo, K. (2004). Information liter-

acy and personal knowledge management. In 12th

Nordic Conference on Information and Documenta-

tion, pages 106–114.

Shneiderman, B. (1989). The Society of Text, chapter Re-

ﬂections on authoring, editing, and managing hyper-

text, pages 115–131. MIT Press, Cambridge, MA,

USA.

Shneiderman, B. (1996). The eyes have it: A task by data

type taxonomy for information visualizations. In VL

’96: Proceedings of the 1996 IEEE Symposium on Vi-

sual Languages, Washington, DC, USA. IEEE Com-

puter Society.

olkel, M. (2010). Personal Knowledge Models with

Semantic Technologies. PhD thesis, Fakult

at f

Wirtschaftswissenschaften, Institut AIFB. To appear.

olkel, M. and Abecker, A. (2008). Cost-beneﬁt analy-

sis for the design of personal knowledge management

systems. In ICEIS 2008 - Proceedings of the Tenth

International Conference on Enterprise Information

Systems, Volume AIDSS, Barcelona, Spain, June 12-

16, 2008, pages 95–105.

olkel, M. and Haller, H. (2006). Conceptual data struc-

tures (CDS) – towards an ontology for semi-formal

articulation of personal knowledge. In Proc. of the

14th International Conference on Conceptual Struc-

tures 2006, Aalborg University - Denmark.

olkel, M. and Schaffert, S., editors (2006). Pro-

ceedings of the First Workshop on Semantic Wikis

at ESWC, Budva, Montenegro., volume 206 of

CEUR Workshop Proceedings ISSN 1613-0073. FZI

Forschungszentrum Informatik Karlsruhe, CEUR-

ws.org. http://CEUR-WS.org/Vol-206 (accessed

06.01.2010).

REQUIREMENTS FOR PERSONAL KNOWLEDGE MANAGEMENT TOOLS

337