CRAFTER: A Persona Generation Tool for Requirements Engineering

Devi Karolita

1,2 a

, John Grundy

1 b

, Tanjila Kanij

1 c

, Humphrey Obie

1 d

and Jennifer McIntosh

3 e

Department of Software Systems and Cybersecurity, Faculty of Information Technology,

Monash University, Melbourne, Australia

Department Informatics Engineering, Faculty of Engineering, Palangka Raya University, Palangka Raya, Indonesia

Faculty of Medicine, Dentistry and Health Sciences, The University of Melbourne, Melbourne, Australia

Keywords:

Requirements Engineering, Persona, Large Language Model, Tool.

Abstract:

Personas, a user characterisation, have been widely used in requirements engineering (RE) to enhance the

understanding of end-users and their needs. However, the persona generation process is time-consuming and

demands familiarity with a user-centered approach. The central issue lies in existing tools for automatically

generating personas, which are restricted to generating persona templates and provide limited user control

to tailor personas according to their speciﬁc needs. This paper introduces CRAFTER, a persona generation

tool that uses Large Language Models (GPT-3.5 model). This tool not only automates persona creation but

also offers recommendations to users for generating personas tailored to their requirements. The study in-

volved an online questionnaire with 19 respondents who utilised the tool, providing feedback that indicated

the tool’s sufﬁciency for persona generation while identifying areas for improvement. Beyond its primary

function, CRAFTER stands out by providing guidance to requirements engineers throughout the persona cre-

ation process. The tool grants users the ﬂexibility to customise personas based on their speciﬁc requirements,

acknowledging the crucial human subjectivity in persona development. Additionally, CRAFTER promotes

persona reusability, allowing users to save and reuse generated personas for future projects.

1 INTRODUCTION

Requirements engineering (RE) tasks involve discov-

ering, deﬁning, and validating end-users’ require-

ments (Sommerville, 2016), emphasising consistent

human interaction to ensure software products meet

users’ needs. Personas, hypothetical archetypes and

descriptive models of real users, complement human

involvement in software engineering (SE) (Cooper,

1999) by representing targeted human beneﬁciaries

(Cooper et al., 2007; Kolski and Warin, 2018). Used

at every RE stage, personas aid requirements engi-

neers in requirements elicitation, identifying diverse

user perspectives (Schneidewind et al., 2012; Mayas

et al., 2016) and discovering new requirements (Ho

and Lin, 2019; Sim and Brouse, 2015; Cleland-Huang

et al., 2013). During requirements speciﬁcation, per-

sonas assist in deﬁning (Sim and Brouse, 2014) and

https://orcid.org/0000-0001-6908-9785

https://orcid.org/0000-0003-4928-7076

https://orcid.org/0000-0002-5293-1718

https://orcid.org/0000-0002-6322-2984

https://orcid.org/0000-0002-6655-0940

Emily is a high school student who comes from a middle-income family

and has two younger siblings. With a passion for art and music, Emily

spends her free time drawing and playing the guitar. She loves being

involved in her school performances.

On a daily basis, Emily relies on Google Maps to explore the city and find

cool art galleries, music stores, and places to hang out with friends.

However, slow loading times and occasional connectivity issues can be

frustrating, especially when she’s in a hurry to meet her friends at a

music event or art exhibition.

Emily, 15 years

Year 10 student

“My vibrant personality shines through my

artistic passion and love for connecting with

friends, making me joyful and creative soul”

Figure 1: Example of a persona.

documenting proposed product requirements (Faily

and Iacob, 2017), while in requirements validation,

they help identify (Aoyama, 2005; Sim and Brouse,

2014) and reﬁne relevant requirements (Araujo and

Aquino Junior, 2014). Figure 1 shows an example of

a persona used in RE-related tasks.

Personas can be automatically generated either us-

674

Karolita, D., Grundy, J., Kanij, T., Obie, H. and McIntosh, J.

CRAFTER: A Persona Generation Tool for Requirements Engineering.

DOI: 10.5220/0012718400003687

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 19th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE 2024), pages 674-683

ISBN: 978-989-758-696-5; ISSN: 2184-4895

ing semi-automatic approaches (Faily and Lyle, 2013;

Alvertis et al., 2016) or fully automated processes

(Jung et al., 2018; An et al., 2018; Kanij et al.,

2023). Large Language Models (LLMs) have also

been utilised in the manual (De Paoli, 2023; Cheng

et al., 2023) and automatic (Zhang et al., 2023) gen-

eration of personas. These approaches have some

shortcomings, such as restricting user choices regard-

ing human facets, providing static attributes, and re-

sulting in reduced user control over persona genera-

tion (Karolita. et al., 2023). We wanted to develop

a tool that employs LLMs to generate personas for

speciﬁc domains while granting users greater control.

Acknowledging the subjective nature of personas, our

tool seeks to ﬁnd a middle ground between automa-

tion and human input, allowing users to customise

personas within predeﬁned domains. This approach

streamlines the persona generation process and is par-

ticularly beneﬁcial for junior requirements engineers

who are still familiarising themselves with the per-

sona concept.

The rest of this paper is organised as follows: in

Section 2 we provide the driving factors of our study

and in Section 3 we explain how we conducted the

study. Section 4 presents the feedback from user eval-

uation as well as presenting the areas for improve-

ment. Section 5 provides a summary of research pa-

pers that are related to our study. Finally, Section 6

concludes the paper.

2 MOTIVATION

Our previous investigation into personas in RE high-

lighted their signiﬁcant role, particularly during the

requirements elicitation and analysis task (Karolita

et al., 2023). Personas proved effective in enhancing

understanding of the needs of software’s human bene-

ﬁciaries. Additionally, many studies underscored the

value of personas in revealing potential interactions

between end-users and the software. This aspect is

crucial for pinpointing and addressing possible chal-

lenges in user-software interactions. This insight into

personas’ utility in RE tasks underscores their impor-

tance in creating software that is closely aligned with

user needs and preferences.

The generation of personas, while beneﬁcial,

presents several challenges. Obtaining representative

data or participants for the target population is often

difﬁcult, posing a signiﬁcant hurdle in creating ac-

curate personas (Lachner et al., 2015; Nielsen et al.,

2013; McIntosh et al., 2021). Furthermore, crafting

personas requires an in-depth understanding of a user-

centric approach in software development (Idoughi

et al., 2012), which can be quite time-consuming

(Acuna et al., 2012; Lopez-Lorca et al., 2014) and po-

tentially expensive to implement effectively (Cleland-

Huang et al., 2013). These challenges highlight the

need for efﬁcient and cost-effective methods in per-

sona creation within the context of software develop-

ment.

Emerging persona generation tools aim to address

challenges in persona creation but come with their

own limitations. Kanij et al.’s tool (Kanij et al., 2023)

focuses on age-speciﬁc personas like those for chil-

dren and older adults, yet it restricts users in inte-

grating diverse human aspects besides age. Similarly,

the ”Automatic Persona Generation” tool (Jung et al.,

2018; An et al., 2018; Jansen et al., 2020), utilising

social media data, is constrained in user control over

persona development. Additionally, ”PersonaGen,”

leveraging Large Language Models (Zhang et al.,

2023), primarily offers limited persona templates,

lacking the depth of more comprehensive persona

proﬁles. These tools indicate progress but also high-

light the need for more versatile and user-controllable

persona generation methods.

Addressing the limitations of existing persona

generation tools, we introduced a novel approach with

our tool, Crafting Recommendations and Advice for

Tailored, Effective Personas (CRAFTER). This tool

is speciﬁcally designed to assist requirements engi-

neers in creating personas tailored to speciﬁc domains

for RE-related tasks. CRAFTER aims to provide

a balanced blend of automation and human input,

enabling users to customise personas to meet their

unique needs within speciﬁc domains. This approach

enhances the user’s control over persona generation,

adding a more personalised touch to the process.

In order to achieve our goal, we wanted to answer

the following key research questions:

RQ1. Can a persona recommendation tool be de-

veloped to generate better personas for use in RE? We

wanted to create a tool that not only helps in the gen-

eration of personas but also proactively offers recom-

mendations for key elements to include in a persona

tailored to a speciﬁc domain.

RQ2. To what extent can such a persona rec-

ommendation tool usable for requirements engineers

generate personas for use in RE? We aimed to assess

our proposed tool’s usability in aiding requirements

engineers with persona generation and provide rec-

ommendations for the tool reﬁnement.

CRAFTER: A Persona Generation Tool for Requirements Engineering

675

3 OUR APPROACH

We developed CRAFTER, a tool designed to auto-

matically generate personas while offering users ﬂex-

ibility in both the persona development process and

the ability to adjust the generated personas. The tool

was developed based on the persona taxonomy and

persona dimensions resulting from our earlier persona

curation study (Karolita. et al., 2023). It incorpo-

rates two major persona layers (internal and exter-

nal), each incorporating human factors (i.e., persona

attributes). This includes the layout of the generated

personas which adhere to the persona dimensions, in-

cluding human factors captured, persona length and

persona narration style. CRAFTER was developed

as a web application utilising a NoSQL database to

record domains of use and persona attributes. Addi-

tionally, we integrated GPT-3.5 model to help in cre-

ating persona descriptions. The tool is publicly acces-

sible

. CRAFTER’s main processes are illustrated in

Figure 2.

3.1 CRAFTER Features

3.1.1 Guided Persona Development

One of CRAFTER’s major features is to generate per-

sonas based on users’ speciﬁc requirements using a

detailed persona taxonomy we formulated in our per-

sona curation study. This detailed persona taxonomy

and persona dimensions resulted from our curation of

98 personas collected from 41 academic publications

(Karolita. et al., 2023). CRAFTER also extends be-

yond persona generation to provide recommendations

to the users about human factors to incorporate into

the personas.

3.1.2 Human-Centred Customisation

A major feature that distinguishes CRAFTER from

other persona development tools is its capacity to of-

fer users a highly tailored persona development ex-

perience. This feature places the power of person-

alisation in the hand of the users, allowing them to

adjust the personas to meet their needs. Users have

the ﬂexibility to choose the domain of implementa-

tion and ﬁne-tune the human factors depicted in the

persona to align with their speciﬁc requirements. This

level of personalisation provides valuable recommen-

dations to users, granting them greater control over

the persona generation process. This feature is es-

pecially advantageous for individuals who are new

to or in the process of familiarising themselves with

http://54.206.127.165:3000/

the concept of personas making the tool more user-

friendly and accessible to a wider audience.

3.1.3 Leveraging a Large Language Models

(LLMs)

To facilitate persona generation, we employed the

GPT-3.5 model in our CRAFTER persona tool. This

enables users to input descriptions about the intended

domain of use and speciﬁc human aspects required

for the personas. The resulting CRAFTER personas

are more highly customised and contextually relevant

to the domain and speciﬁed human characteristics of

interest. The use of the LLM ensures consistency

in descriptions and is scalable, accommodating sin-

gle or multiple personas for diverse scenarios. Its

user-friendliness reduces the skill barrier for persona

creation, making it accessible to a broader audience.

Leveraging a very large language model with deep

information about many software domains and hu-

man characteristics, CRAFTER provides high accu-

racy and high-quality persona descriptions.

3.1.4 Persona Reusability

CRAFTER includes a feature allowing users to save

the generated personas for reﬁnement and reuse.

These saved personas can be readily reused or ad-

justed in accordance with the users’ requirements.

This functionality offers users the ﬂexibility to work

with previously developed personas, saving them time

and effort in the persona creation process.

3.2 CRAFTER’s Persona Taxonomy

We developed CRAFTER based on our detailed per-

sona taxonomy and persona dimensions (Karolita.

et al., 2023): we developed a Persona Corpus, then

mapped the persona attributes (i.e., human factors in-

cluded in the personas) and divided them into two

layers: the internal layer and the external layer (see

Table 1). The internal layer of a persona consists

of general background information including personal

characteristics and the external layer contains context-

speciﬁc information based on the context and/or the

domain in which the personas are used.

As personas are context-speciﬁc tools, we know

that a persona used in a particular domain requires

some customisation in terms of attributes for that do-

main and might not be applicable to other domains.

Therefore, we have also recommended persona at-

tributes for each domain to address the speciﬁc re-

quirements. We also discovered multiple persona

styles (referred to as persona dimensions), in which

text-based personas can be presented. Our Persona

ENASE 2024 - 19th International Conference on Evaluation of Novel Approaches to Software Engineering

676

PERSONA

TAXONOMY

DATABASE DESIGN INTERFACE DESIGN

LARGE LANGUAGE

MODELS

GENERATED

PERSONAS

Figure 2: The main processes of CRAFTER.

Table 1: Persona taxonomy (from (Karolita. et al., 2023)).

INTERNAL LAYER

EXTERNAL LAYER

Personal characteristics

Demographic information

Personal attributes

Motivation

Goal

Concern/frustration/pain point

Skill/experiential/environ

mental-influenced

characteristics

Personal story

Interaction with technology

Group or multiple human

characteristics

Work status

Family environment

Geographic location

Collaboration and

communication style

Corpus identiﬁed three key dimensions for represent-

ing text-based personas: persona narration, persona

format, and persona length.

3.3 CRAFTER Example Usage

When users access CRAFTER for persona genera-

tion, they must either sign in (for new users) or log

in with their existing account. This requirement is in

place because CRAFTER offers users the ﬂexibility

to either initiate the persona generation process or re-

view their previously saved personas, as depicted in

Figure 3. This functionality ensures a personalised

experience, allowing users to access and manage their

unique persona creations efﬁciently.

In the persona generation process with

CRAFTER, users begin by selecting the domain

of use for their persona, as shown in Figure 4. The

tool offers a range of predeﬁned domains, but users

also have the ﬂexibility to add new ones as required.

Next, users can reﬁne their persona by specifying

various human factors, which include both internal

and external elements. There’s also an option to

introduce new human factors, enhancing the tool’s

versatility. This feature makes CRAFTER dynamic,

catering to diverse and speciﬁc persona development

needs.

Figure 5 shows the subsequent steps in

CRAFTER, where users can choose their de-

sired persona length and narrative style. Regarding

persona length, users have the option to select the

word range for the generated persona. Additionally,

users can specify the persona’s narration style, choos-

ing between a narrative format or a bullet-points

style. These user-deﬁned persona requirements (i.e.,

the domain of use, human factors, persona length,

and narration style) serve as a prompt GPT-3.5 to

generate the persona. An example of the gener-

ated persona is presented in Figure 6. Users can

re-generate the persona and once users are satisﬁed

with the generated personas, they can save them for

future access.

4 EVALUATION

The CRAFTER tool underwent a usability evaluation

involving the System Usability Scale (SUS) (Brooke,

1995) and qualitative feedback to assess its effective-

ness. The evaluation was conducted through an online

questionnaire

, which had received ethical approval

from the Monash University Human Research Ethics

Committee (Approval #38469), ensuring adherence to

ethical research standards. This thorough assessment

aimed to gather insightful feedback on CRAFTER’s

usability and overall user experience in the context of

persona generation for RE tasks.

4.1 Participants

Nineteen respondents answered our questionnaire,

which included demographics, the SUS scale, and

follow-up in-depth questions. The respondents’ cur-

rent roles included Software Practitioner, Require-

ments Engineer, UX Practitioner, Software Archi-

tect, Artiﬁcial Intelligence (AI) Practitioner, and Aca-

demic; some holding multiple roles. The major-

ity were male and aged between 26 to 34 years.

They were predominantly university-educated, with

degrees in Software Engineering or Computer Sci-

ence (Figure 7). The majority of the respondents were

familiar with persona concepts and had designed one

or more personas. They also incorporated personas

in RE-related tasks (such as eliciting, specifying, and

https://doi.org/10.5281/zenodo.10662689

CRAFTER: A Persona Generation Tool for Requirements Engineering

677

Figure 3: CRAFTER home page.

Figure 4: Customising a persona’s domain of use and human factors.

validating user requirements). Figure 8 summarises

the persona’s domain of use and how the respondents

used personas in their project.

4.2 User Feedback

In our questionnaire, we calculated the results us-

ing the System Usability Scale (SUS) methodology

(Brooke, 1995). The SUS scores for each participant,

identiﬁed as R-i for the ith respondent, are presented

in Table 2. This table includes the Score-i, which is

the total SUS score assigned to each respondent, and

the Final Score-i, representing the respondent’s ulti-

mate SUS score. This approach allows us to quan-

titatively assess the usability of CRAFTER based on

participants’ feedback.

In the evaluation of CRAFTER, we categorised

the ﬁnal SUS scores of the respondents into three lev-

els of acceptance as per Bangor et al.’s criteria (Ban-

gor et al., 2009): Not acceptable, Marginal, and Ac-

ceptable. Out of the respondents, six rated CRAFTER

as ”Acceptable”, ﬁve found it ”Marginal”, and eight

considered it ”Not acceptable”. This categorisation,

based on the SUS methodology, provides insight into

the users’ perceptions of CRAFTER’s usability. The

distribution of these scores, depicted in a Figure 9,

offers a clear understanding of how the tool fares in

terms of user satisfaction and acceptability.

Positive feedback from respondents regarding

CRAFTER includes comments on its straightforward

nature. One respondent commended the tool for be-

ing easy to use, while another acknowledged its util-

ity but suggested a learning curve. An unexpected

positive reaction came from a respondent who was

pleasantly surprised by the tool’s ability to generate

complete personas. Two respondents emphasised the

essential features offered by CRAFTER, highlighting

its role in streamlining the persona generation process

and eliminating the need to start from scratch. Addi-

tionally, a respondent expressed satisfaction with the

tool’s ﬂexibility, appreciating the ability to customise

personas according to speciﬁc needs.

We identiﬁed several limitations in CRAFTER

based on the respondents’ feedback which were cate-

ENASE 2024 - 19th International Conference on Evaluation of Novel Approaches to Software Engineering

678

Figure 5: Selecting preferred persona style.

Table 2: SUS individual results.

R-i Score-i Final Score-i R-i Score-i Final Score-i

R1 19 47.50 R11 35 87.50

R2 26 65 R12 19 47.50

R3 27 67.50 R13 39 97.50

R4 16 40 R14 20 50

R5 23 57.50 R15 19 47.50

R6 7 17.50 R16 12 30

R7 17 42.50 R17 29 72.50

R8 34 85 R18 14 35

R9 31 77.50 R19 33 82.50

R10 23 57.50

Areef is 70 years old and is retired. After retirement, he found

that he has too much free time and he used to feel bored.

Sometimes, he tutors primary school children voluntarily as

he was a primary school teacher before.

His friends are also mostly retired and he does not want to

disturbs them during their working hours. He also has

children but they are staying quite far from his place.

He love to socialise with other older adults through

community centres but due to his physical challenge he could

not frequent these centres as often as he wish.

Lately he found out that he enjoys using smartphone,

especially on discovering news, and he thinks that by using

smartphone he does not always need to go to community

centres to get updates about activities that he is interested

in.

Figure 6: An example of persona generated by CRAFTER.

gorised into two major groups: (1) limitations related

to the developed personas and (2) limitations related

to the tool.

Limitations related to the generated personas in-

cluded some of our respondents noting that the per-

sona layout does not resemble the actual personas

used in RE practices. They also found that the gen-

erated personas sometimes lacked speciﬁcity and did

not align with their requirements.

Limitations related to the tool were identiﬁed as

due to user experience problems, including ﬁndabil-

ity, affordance, and interaction cost. Findability en-

compasses issues where users struggle to locate infor-

mation or navigate within the tool. Many of the lim-

itations associated with the tool were centred around

ﬁndability, where Users found it challenging to access

the necessary information. For instance, our respon-

dents expressed difﬁculties in understanding the tool’s

purpose and how it aids in persona generation.

There were comments about the lack of sufﬁcient

information about the speciﬁc details to be provided

for the generated persona. We also identiﬁed lim-

itations related to interaction cost, where users had

to engage in excessive interactions to perform cer-

tain tasks. For example, they needed to scroll ex-

cessively to access certain persona attributes. Re-

spondents mentioned difﬁculties in navigation and ex-

pressed confusion about the steps required to gener-

ated personas. Limitations related to affordance in-

dicate a lack of guidance for users to interact with

the interface and anticipate what to expect. For in-

stance, users were unsure whether the generated per-

sonas were saved or not.

4.3 Areas for Improvement

Based on the user feedback, we have identiﬁed sev-

eral areas for improvement to enhance our tool (sum-

marised in Table 3). To address limitations related

to the tool itself, we aim to reﬁne the user interface

to offer a more intuitive experience. This involves

facilitating smoother navigation and interaction, in-

corporating features such as a ﬁlter function to pre-

vent excessive scrolling, providing an option for one-

off users to generate personas without creating an ac-

count, and clearly describing each task required to

generate a persona. Moreover, we recognise the im-

CRAFTER: A Persona Generation Tool for Requirements Engineering

679

Figure 7: Questionnaire respondents’ demographics.

Figure 8: Questionnaire respondents’ persona usage.

portance of offering comprehensive information about

persona concepts, and their signiﬁcance in RE-related

tasks, and especially targeting users who may be new

to or unfamiliar with persona concepts. A prelim-

inary introduction detailing the tool’s purpose, fea-

tures, and functionalities can guide users more effec-

tively through the persona generation process.

One suggestion for improving CRAFTER in-

cludes integrating the latest Large Language Models

(LLMs) like GPT-4 for more accurate persona gener-

ation. This could result in personas that are closely

aligned with users’ speciﬁc needs and requirements.

Additionally, enhancing the tool with options to cus-

tomise the layout of generated personas would allow

users to tailor the design according to their prefer-

ences and team standards, potentially increasing the

tool’s usability and relevance across different scenar-

ios. We aim to make CRAFTER more adaptable

to diverse domains and contexts of persona use. To

achieve this, we plan to allow users to specify not only

the domain of use but also the context of use for the

persona. For example, in the health domain, a user

might want to create a persona for a patient, health

worker, or caregiver. This will allow a broader range

of users to harness the tool’s capabilities effectively.

We are considering the implementation of collabo-

rative features that enable multiple users to work to-

gether in generating personas. This supports persona

usage teamwork and knowledge sharing, ultimately

enhancing the persona generation process.

5 RELATED WORK

In practice, most personas are predominantly gener-

ated through manual methods (Karolita et al., 2023).

Three primary techniques are commonly employed:

qualitative, quantitative, and mixed methods tech-

niques (Tu et al., 2010; Jansen et al., 2021; Jansen

et al., 2022). Qualitative methods represent the most

prevalent approach, involving activities like inter-

views, focus group discussions, brainstorming, and

workshops. Mixed-methods approaches to persona

creation generally begin with qualitative techniques

in the initial phase of the process. This includes

methods such as interviews, observations, and ﬁeld

studies, followed by the application of quantitative

techniques for data analysis (Mesgari et al., 2019;

ENASE 2024 - 19th International Conference on Evaluation of Novel Approaches to Software Engineering

680

SUS score

0 5 10 15 20

Not acceptable Marginal - Low Marginal - High Acceptable

Figure 9: CRAFTER’s acceptability ranges.

Table 3: Areas for improvement.

Tool Generated personas

Better user interface Flexibility to change persona layout

Use the latest advancement of LLM Adaptability to diverse contexts or domains of use

Better tool navigation

Provide preliminary introduction on how to navigate the tool

Provide information about persona and why it matters

Enabling collaborative persona creation

Mead et al., 2017). Additionally, personas can be

generated through quantitative methods by either us-

ing secondary sources (e.g., forum posts, user re-

views) (Rahimi and Cleland-Huang, 2014) or em-

ploying questionnaires to gather substantial amounts

of data for persona generation (Schafer et al., 2019).

This approach involves data mining techniques such

as cluster analysis.

Manual persona generation in Requirements Engi-

neering, although essential, faces several challenges.

Firstly, accessing representative participants or data

for the target population is often difﬁcult, which can

affect the accuracy and relevance of the personas

(Lachner et al., 2015; McIntosh et al., 2021). Fur-

thermore, creating effective personas necessitates a

deep understanding of user-centred design principles

(Idoughi et al., 2012). This process is notably time-

consuming, involving various tasks that extend the

overall timeline of RE activities (Acuna et al., 2012;

Lopez-Lorca et al., 2014). Consequently, it can lead

to increased project durations and higher budget re-

quirements (Cleland-Huang et al., 2013), which may

be challenging for resource-constrained projects.

Consequently, several studies have explored the

automation of persona generation to mitigate the

aforementioned challenges associated with manual

persona creation. For instance, Faily and Lyle (Faily

and Lyle, 2013) integrated Computer Aided Inte-

gration of Requirements and Information Security

(CAIRIS)

with a Persona Case framework (Faily

and Flechais, 2011). They employed a framework to

qualitatively analyse collected data, resulting in the

identiﬁcation of new persona characteristics. CAIRIS

was then utilised to augment existing personas with

https://github.com/cairis-platform/cairis

the newly identiﬁed persona characteristics, thus gen-

erating new personas. Alvertis et al. (Alvertis et al.,

2016) employed a crowd-sourcing technique, Per-

sona Builder, to create a pool of user proﬁles, which

also stores previously generated personas. To create

personas for a proposed product, software develop-

ers utilise a computer-aided tool, known as Persona

Builder. They selected persona characteristics from

the pool or reused existing personas, and the tool au-

tomatically amalgamated the chosen characteristics to

generate personas accordingly.

Several studies have contributed to the ﬁeld of au-

tomated persona generation. One such initiative is

the Automatic Persona Generation tool, which creates

personas based on online behavioral data of social

media users (Jung et al., 2018; An et al., 2018; Jansen

et al., 2020). It employs quantitative methods, utilis-

ing users’ demographic information and their online

interactions. Additionally, Kanij et al. have devel-

oped a tool that speciﬁcally generates personas for

certain age groups, such as children and older adults,

based on age-related characteristics identiﬁed through

a systematic literature review (Kanij et al., 2023). An-

other innovative approach involves using Large Lan-

guage Models (LLMs), particularly the GPT-4 model,

along with a knowledge graph to automatically gen-

erate persona templates, incorporating user feedback

in the process (Zhang et al., 2023). Although these

studies claim that their tools expedite persona gener-

ation, they have limitations, such as offering limited

options for including diverse human facets (beyond

age) (Kanij et al., 2023), and constraints in provid-

ing user control during the persona generation process

(Jung et al., 2018; An et al., 2018; Jansen et al., 2020;

Zhang et al., 2023).

CRAFTER: A Persona Generation Tool for Requirements Engineering

681

6 SUMMARY

This paper introduces CRAFTER, an innovative tool

designed to automate persona generation for RE prac-

tices. This tool uniquely combines the principles

of persona taxonomy and advanced technologies like

Large Language Models (LLM), facilitating the cre-

ation of high-quality, nuanced personas. CRAFTER

stands out for its user-centric approach, providing re-

quirements engineers with the ability to tailor per-

sonas to their speciﬁc project needs. This func-

tionality is particularly advantageous for those new

to persona development, offering a guided and cus-

tomisable experience. In addition to its primary role,

CRAFTER supports the reuse of personas, enhancing

efﬁciency in future projects. To evaluate CRAFTER’s

effectiveness, we conducted an extensive question-

naire with 19 respondents, which yielded generally

positive feedback and identiﬁed key areas for im-

provement. These insights will be instrumental in the

ongoing development of CRAFTER, focusing on re-

ﬁning both the user interface and the depth of persona

details generated. The study highlights CRAFTER’s

potential as a tool that not only automates but also

enriches the persona creation process in RE. Future

work is aimed at enhancing CRAFTER’s capabilities

based on user feedback, ensuring it meets the evolv-

ing needs of requirements engineers.

ACKNOWLEDGEMENTS

Karolita is supported by Australia Awards Scholar-

ship and Monash Departmental Top-Up Scholarship

for her Ph.D. study at Monash University, Australia.

Grundy, Kanij, and McIntosh are supported by the

Australian Research Council (ARC) Laureate Fellow-

ship project FL190100035. McIntosh is also funded

by a National Health and Medical Research Coun-

cil (NHMRC) Synergy Grant (APP2010268) and

NHMRC Participation in Cancer Screening Programs

Grant (APP2014703).

REFERENCES

Acuna, S. T., Castro, J. W., and Juristo, N. (2012). A hci

technique for improving requirements elicitation. In-

formation and Software Technology, 54(12).

Alvertis, I., Papaspyros, D., Koussouris, S., Mouzakitis, S.,

and Askounis, D. (2016). Using crowdsourced and

anonymized personas in the requirements elicitation

and software development phases of software engi-

neering. In 2016 11th Int. Conf. on Availability, Reli-

ability and Security (ARES). IEEE.

An, J., Kwak, H., Jung, S., Salminen, J., Admad, M., and

Jansen, B. (2018). Imaginary people representing real

numbers: Generating personas from online social me-

dia data. ACM transactions on the web, 12(4):1–26.

Aoyama, M. (2005). Persona-and-scenario based require-

ments engineering for software embedded in digital

consumer products. In 13th IEEE Int. Conf. on Re-

quirements Engineering (RE’05). IEEE.

Araujo, C. F. d. and Aquino Junior, P. T. (2014). Psycholog-

ical personas for universal user modeling in human-

computer interaction. In International Conference on

Human-Computer Interaction, pages 3–13. Springer.

Bangor, A., Kortum, P., and Miller, J. (2009). Determining

what individual sus scores mean: Adding an adjective

rating scale. J. Usability Studies, 4(3):114–123.

Brooke, J. (1995). Sus: A quick and dirty usability scale.

Usability Eval. Ind., 189.

Cheng, M., Durmus, E., and Jurafsky, D. (2023). Marked

personas: Using natural language prompts to measure

stereotypes in language models. arXiv.org.

Cleland-Huang, J., Czauderna, A., and Keenan, E. (2013).

A persona-based approach for exploring architec-

turally signiﬁcant requirements in agile projects. In

Int. Working Conf. on Requirements Engineering:

Foundation for Software Quality. Springer.

Cooper, A. (1999). The inmates are running the asylum.

Software-Ergonomie ’99.

Cooper, A., Reimann, R., and Cronin, D. (2007). About

Face 3 : The Essentials of Interaction Design. Wiley,

Newark, UNITED STATES.

De Paoli, S. (2023). Writing user personas with large lan-

guage models: Testing phase 6 of a thematic analysis

of semi-structured interviews. arXiv.org.

Faily, S. and Flechais, I. (2011). Persona cases: a technique

for grounding personas. In Conference on Human

Factors in Computing Systems - Proceedings, CHI

’11, pages 2267–2270. ACM.

Faily, S. and Iacob, C. (2017). Design as code: Facil-

itating collaboration between usability and security

engineers using cairis. In 2017 IEEE 25th Interna-

tional Requirements Engineering Conference Work-

shops (REW), pages 76–82.

Faily, S. and Lyle, J. (2013). Guidelines for integrating

personas into software engineering tools. In Proceed-

ings of the 5th ACM SIGCHI Symposium on Engineer-

ing Interactive Computing Systems, EICS ’13, page

69–74, New York, NY, USA. Association for Com-

puting Machinery.

Ho, S.-H. and Lin, C. J. (2019). The requirement analysis

for developing the assisted living technology for the

elderly. In Int. Conf. Cognitive Cities.

Idoughi, D., Seffah, A., and Kolski, C. (2012). Adding user

experience into the interactive service design loop: a

persona-based approach. Behaviour & Information

Technology, 31(3):287–303.

Jansen, B. J., Jung, S.-G., and Salminen, J. (2020). From ﬂat

ﬁle to interface: Synthesis of personas and analytics

for enhanced user understanding. Proceedings of the

Association for Information Science and Technology,

57(1):e215.

ENASE 2024 - 19th International Conference on Evaluation of Novel Approaches to Software Engineering

682

Jansen, B. J., Soon-Gyo, J., Nielsen, L., Guan, K. W., and

Salminen, J. (2022). How to create personas: Three

persona creation methodologies with implications for

practical employment. Paciﬁc Asia Journal of the As-

sociation for Information Systems, 14(3):1.

Jansen, J., Jung, S.-G., Salminen, J., Guan, K., and Nielsen,

L. (2021). Strengths and weaknesses of persona cre-

ation methods: Guidelines and opportunities for digi-

tal innovations.

Jung, S.-g., Salminen, J., Kwak, H., An, J., and Jansen,

B. J. (2018). Automatic persona generation (apg) a

rationale and demonstration. In Proceedings of the

2018 conference on human information interaction &

retrieval, pages 321–324.

Kanij, T., Du, X., Grundy, J., Madugalla, A., and Karolita,

D. (2023). An approach to generating diverse per-

sonas for children and the elderly for software de-

velopment. In 2023 IEEE 47th Annual Computers,

Software, and Applications Conference (COMPSAC),

pages 898–903. IEEE.

Karolita., D., Grundy., J., Kanij., T., Obie., H., and McIn-

tosh., J. (2023). What’s in a persona? a preliminary

taxonomy from persona use in requirements engineer-

ing. In ENASE2023, pages 39–51.

Karolita, D., McIntosh, J., Kanij, T., Grundy, J., and Obie,

H. O. (2023). Use of personas in requirements en-

gineering: A systematic mapping study. Information

and Software Technology, 162:107264.

Kolski, C. and Warin, B. (2018). From Persona to Liv-

ing Persona, Preliminary Data from a Pilot Study in

HCI Education. Int. Conf. Learning and Collabora-

tion Technologies.

Lachner, F., von Saucken, C., ‘Floyd’ Mueller, F., and Lin-

demann, U. (2015). Cross-cultural user experience

design helping product designers to consider cultural

differences. In Cross-Cultural Design Methods, Prac-

tice and Impact, Lecture Notes in Computer Science.

Lopez-Lorca, A. A., Miller, T., Pedell, S., Mendoza, A.,

Keirnan, A., and Sterling, L. (2014). One size doesn’t

ﬁt all: diversifying” the user” using personas and emo-

tional scenarios. In 6th Int. Workshop on Social Soft-

ware Engineering.

Mayas, C., H

orold, S., and Kr

omker, H. (2016). Personas

for requirements engineering: Opportunities and chal-

lenges. In Usability-and Accessibility-Focused Re-

quirements Engineering: First International Work-

shop, UsARE 2012, Held in Conjunction with ICSE

2012, Zurich, Switzerland, June 4, 2012 and Second

International Workshop, UsARE 2014, Held in Con-

junction with RE 2014, Karlskrona, Sweden, August

25, 2014, Revised Selected Papers 1, pages 34–46.

Springer.

McIntosh, J., Du, X., Wu, Z., Truong, G., Ly, Q., How,

R., Viswanathan, S., and Kanij, T. (2021). Evaluat-

ing age bias in e-commerce. In 2021 IEEE/ACM 13th

International Workshop on Cooperative and Human

Aspects of Software Engineering (CHASE), pages 31–

40. IEEE.

Mead, N., Shull, F., Spears, J., Heibl, S., Weber, S., and

Cleland-Huang, J. (2017). Crowd sourcing the cre-

ation of personae non gratae for requirements-phase

threat modeling. In 2017 IEEE 25th International Re-

quirements Engineering Conference (RE), 4-8 Sept.

2017, 2017 IEEE 25th International Requirements En-

gineering Conference (RE). Proceedings, pages 412–

17. IEEE Computer Society.

Mesgari, M., Okoli, C., and Guinea, A. O. d. (2019). Creat-

ing rich and representative personas by discovering af-

fordances. IEEE Transactions on Software Engineer-

ing, 45(10):967–983.

Nielsen, L., Nielsen, K. S., Stage, J., and Billestrup, J.

(2013). Going global with personas. In Human-

Computer Interaction – INTERACT 2013, volume

8120 of Lecture Notes in Computer Science, pages

350–357, Berlin, Heidelberg. Springer Berlin Heidel-

berg.

Rahimi, M. and Cleland-Huang, J. (2014). Personas in the

middle: automated support for creating personas as

focal points in feature gathering forums. In Proceed-

ings of the 29th ACM/IEEE international conference

on automated software engineering, pages 479–484.

ACM.

Schafer, K., Rasche, P., Brohl, C., Theis, S., Barton, L.,

Brandl, C., Wille, M., Nitsch, V., and Mertens, A.

(2019). Survey-based personas for a target-group-

speciﬁc consideration of elderly end users of infor-

mation and communication systems in the german

health-care sector. International Journal of Medical

Informatics, 132.

Schneidewind, L., H

orold, S., Mayas, C., Kr

omker, H.,

Falke, S., and Pucklitsch, T. (2012). How personas

support requirements engineering. In 2012 First In-

ternational Workshop on Usability and Accessibility

Focused Requirements Engineering (UsARE), pages

1–5. IEEE.

Sim, W. W. and Brouse, P. (2015). Developing ontologies

and persona to support and enhance requirements en-

gineering activities – a case study. Procedia computer

science, 44.

Sim, W. W. and Brouse, P. S. (2014). Empowering require-

ments engineering activities with personas. volume 28

of Procedia Computer Science.

Sommerville, I. (2016). Software Engineering, Global Edi-

tion.

Tu, N., He, Q., Zhang, T., Zhang, H., Li, Y., Xu, H., and Xi-

ang, Y. (2010). Combine qualitative and quantitative

methods to create persona. In 2010 3rd International

Conference on Information Management, Innovation

Management and Industrial Engineering, volume 3,

pages 597–603. IEEE.

Zhang, X., Liu, L., Wang, Y., Liu, X., Wang, H., Ren, A.,

and Arora, C. (2023). Personagen: A tool for gen-

erating personas from user feedback. In 2023 IEEE

31st International Requirements Engineering Confer-

ence (RE), pages 353–354. IEEE.

CRAFTER: A Persona Generation Tool for Requirements Engineering

683