Chatbot in a Campus Environment: Design

of LiSA, a Virtual Assistant to Help

Students in Their University Life

Massimiliano Dibitonto, Katarzyna Leszczynska,

Federica Tazzi ✉ , and Carlo M. Medaglia
( )

DASIC, Link Campus University, Via del Casale di S. Pio V, 44, 00165 Rome, Italy
{m.dibitonto,k.leszczynska,f.tazzi,c.medaglia}@unilink.it

Abstract. This work presents some initial results of our research about the design
and implementation of “LiSA” (Link Student Assistant), a chatbot intended to
help students in their campus life, through information and services. The focus
of our research is to understand which kind of information and services are better
accessed through this kind of touch point, how the chatbot personality influences
the user experience and the interaction and which level of intelligence should be
implemented. After an analysis of the state of the art in the considered application
domain we investigated, through a survey, the users’ needs and their inclination
to the use of a chatbot for this specific purpose. A chatbot was created to deliver
the survey, allowing to understand both the users’ needs and their behaviour while
using the tool.

Keywords: Chatbot · Virtual assistant · Human computer interaction ·

Conversational interfaces · User experience

1 Introduction

Conversational interfaces are not new, however nowadays chatbots and virtual assistants
are spreading on a wide scale as touch points to access information and services in a
more natural way. This process of coming back to textual interaction was speeded up
by the spreading of messaging platforms that started to support chatbots (i.e. Facebook,
Telegram and Whatsapp) and many companies caught the opportunity to establish a new
communication channel with the users. As an example Mastercard allows developers to
easily integrate Masterpass payment technology into conversational e-commerce expe‐
riences through the Facebook chat [1].
Chatbots are used as a channel to interact with a wide range of different “intelligence”
levels, from simple answering machines to complex natural language processors and
artificial intelligence algorithms. Asking something via a text message could seem easier
than browsing a GUI searching for a button or a textbox, however this kind of interfaces
bring out new challenges in terms of user experience, interaction design and usability.
Moreover there is the need of a multidisciplinary approach including disciplines like
psychology, sociology and sociolinguistics.

© Springer International Publishing AG, part of Springer Nature 2018

M. Kurosu (Ed.): HCI 2018, LNCS 10903, pp. 103–116, 2018.
https://doi.org/10.1007/978-3-319-91250-9_9
104 M. Dibitonto et al.

To ensure a “natural” interaction a “good” chatbot should create an “illusion of life”

[2], merging an appearance of intelligence and empathic understanding of the user.
Chatbots, as social agents, need a mind, a body and a personality in order to be believable.
Personality is essential to shape an empathic relationship with the user and to be able to
react to unexpected situations, like verbal aggression, misspellings or lacks in its knowl‐
edge database. In relation with the field of use and its scope (i.e. a financial consultant
in an home banking system, a customer care assistant) it should have a different tone of
voice and offer to the user a different set of functionalities.
The final scope of this work is to investigate how to design a chatbot that could help
students in their campus life. We are using a User Centered Design approach with a
specific focus on interaction and user experience aspects. An important point is related
about finding out which kind of services and information are better delivered with a
chatbot according to users’ needs, users’ perception, context of use and usability. The
chatbot, for example, could provide general information or give personalized messages
(thanks to the knowledge of students’ curriculum). But it could also provide transactional
services, like booking an exam or paying the annual fee. In this case would the student
be more comfortable using a GUI or a chatbot? Furthermore the messages could be
delivered in pull and/or push mode. In the first case the student asks for something, in
the second one the system, based on some parameters, decides to write to the user.
Another important point is related on how the chatbot personality and the tone of voice
influences the User Experience and which one is more suitable for the context of use.
Should the chatbot have the personality of a tutor, a professor, a member of the admin‐
istrative staff? As discussed in the analysis of the state of the art, personality has a strong
impact on the system. Furthermore the level of Artificial Intelligence should be decided
according to the functionalities required but also considering the costs.
In this paper we present the initial steps of our work consisting in a review of the
state of the art and the results achieved during the user research.

2 Background

2.1 Chatbots: A Short Overview

Chatbots are machine agents that engage a communication with human using natural
language. From the early stages, when chatbots were mainly experimented to fool people
into thinking that they were humans and to entertain them (the most famous examples
are ELIZA [3] or A.L.I.C.E. [4]), nowadays they are used into a wide range of fields for
more practical purposes, like information retrieval, e-commerce, education, etc. Over
the time the main focus in chatbot design moved from the perfect imitation of human to
the development of helpful tools able to support the users in their tasks with a more
natural interaction achieved through the use of natural language [5]. This was possible
thanks to the growth of data-mining and machine-learning techniques from one side,
and to the spread of mobile internet and mobile messaging applications from the other
side, that have influenced users’ approach to customer services and technology [6].
Internet users prefer to use social media for asking assistance instead of calling or writing
an email expecting a real-time response [7]. From this point of view chatbots could be
 Chatbot in a Campus Environment: Design of LiSA 105

an effective alternative to traditional customer service thanks to their timely and efficient
assistance and information, which saves the users’ time and made easier and faster obtain
help or information [8].
Different types of virtual assistants can be classified by various criteria. From the
technological point of view Dialog Systems can be divided in two main groups. High
Level Dialog System, that provides a high level of analytical complexity and of context
understanding, and Low Level Dialog System, simple pattern-matching algorithm, with
low level of analytical complexity and a basic context understanding (this systems mimic
the conversation rather than understanding it, but are programmed with a large store of
small talk canned responses that allow to entertain different conversational topic
settings) [9]. This first distinction is closely connected with another important classifi‐
cation between General Personal Assistant (like Apple’s Siri, Amazon’s Alexa, etc.)
and Specialized Digital Assistant (this kind of chatbots has increased significantly after
the launch of chatbots platforms by Facebook, Skype, Telegram, etc. and are more
focused on specific domains or task, like buying a flight, asking information to a specific
customer care or use a learning platform) [10].
The communication with a chatbot can take place by writing text or with a speech-
recognition system, and machine learning methods can be used to train the chatbot to
understand the inputs provided by the users. From here the split between two different
kind of inputs and outputs, closely linked to the context of use. Voice-driven voicebots,
based on speech recognition technology (digital assistants like Apple’s Siri, Amazon’s
Alexa, etc.) that could be consulted everytime, also while the user is driving a car, but
it may have trouble recognizing words if not pronounced correctly. Text-based chatbots,
real time chats, implemented on the website or provided by social media platforms.
There are also virtual assistant that can use both technologies.
If analyzed from the interaction point of view the human-system dialogue is
composed by a person (user) looking for information, and an expert (system) giving
information. This can take place through three interaction strategies, that determinate
who between the user, the system or both are leading the conversation. In case of System-
initiated interaction users’ inputs are limited but the interaction is more efficient, in
reverse in a User-initiated interaction users are free to take initiative but there can be
problem with understand all the users’ inputs. Mixed-initiated interaction strategy seems
to be the solution better adaptable to different contexts [11].

2.2 Conversational Interfaces and User Experience

Understanding the user is fundamental for designing better conversational interfaces
and more efficient chatbots, also because users expect to use natural language in their
conversations with this kind of technology. In this scenario user’s digital interaction
moved from graphical user interfaces (websites and apps) to messaging platforms and
voice-driven digital assistants [12]. Design should not be seen as an explanatory task
(the user can see all the content and features available to reach his goal) but as an inter‐
pretational task (the system must understand the users’ needs and how to serve him) and
must extends from the design of the UI to the whole service [6]. The user interface is
no longer limited to the screens but embraces other forms of interaction, above all the
106 M. Dibitonto et al.

machines are starting to understand user’s natural words, behaviors and gestures. It’s
no longer necessary for the user to understand the machine’s language (command line
or GUI). About this change Andy Goodman, group director of Fjord, introduces the
provocative term of Zero UI [13].
An important change in the sociology of the users of computer applications is the
transition from expert users to “novice users”, who need a virtual assistant more user-
friendly, that satisfies some requirements in order to fulfill their requests. Sansonnet,
Leray and Martin have identified a basic framework that a virtual assistant must observe.
It must understand the user through the function of comprehension (The Dialogical
Agent), must answer to the user’s question trough the function of competence and an
access to a database of knowledge storage (The Rational Agent) and must conquer the
trust of the user through the function of presence (The Embodied Agent) [14].
Kuligowska et al. proposed an interesting framework evaluation method to identify
different aspects of chatbots functioning, which includes Visual Look, Knowledge Base
(general and specialized information because a conversational agent should be able to
answer to a set of general knowledge questions, not only about the topic in which it is
specialized, an aspect close to the personality traits of the chatbot), Conversational
Abilities, Language Skills and Context Sensitiveness (a good chatbot should be able to
bear a coherent dialogue, understand the context, giving feedback and repairing the
dialogue if necessary, to minimize user frustration during the interaction), Personality
Traits (a successful chatbot should have a unique personality, including personality
aspects, biography and emotions), Emergency Responses in Unexpected Situations (the
chatbot should be able to overcome user’s misspellings or other mistakes and diplo‐
matically answers to rude or sexual provocations), Possibility of Rating Chatbot (ask
directly to the users to rate the conversation is an important added value to estimate the
appreciation of the virtual agent) [15].
This frameworks underlines the need to integrate the technology with psychosocial
factors and to include the users’ linguistic, cognitive and social behaviour. The effec‐
tiveness, efficiency and satisfaction of the classic usability definition is not enough in
this case. Chatbots must establish a relationship with users, must be appealing and
entertaining. They have to be flexible (user’s personalisation and bot’s adaptivity to the
users’ model), affective (the quality of the relationship), communicative (minimum
effort required for the interaction) and autonomous (not only reaction to users’
command). They need a mind, a body and a personality to be able to be successful and
believable. The mind consists in cognitive abilities (problem solving), social capabilities
and affective sensitivity (understand the mood of the user and answer with the correct
emotional state). The body can be visible or not, for example described by a good narra‐
tive. The personality defines the interaction style and the “character” of the bot. The
combination of this three aspects generate the bot’s behaviour [2].
Before designing a chatbot it is important to know the users’ motivations for using
it. The user actively chooses the best medium for his purpose and an emotionally engage
with him could be the variable that could make him choose to use the chatbot. Virtual
assistant doesn’t really “understand” the dialogue but they can create the illusion of
intelligence and empathic understanding of the user and must be able to react to unex‐
pected situations, like verbal aggression, misspellings or lacks in its knowledge database.
 Chatbot in a Campus Environment: Design of LiSA 107

The goal of the “illusion of life” requires a multidisciplinary approach from artistic,
psychologic and technologic fields [2]. For example the use of participant’s first name
during a conversation with a chatbot is perceived as positive and influence positively
the all conversation [16]. It could be also useful consider additional contextual infor‐
mation like users’ social media profiles to better empathize with their emotional status,
even because the users’ requests on social media can be both informational and
emotional [7].

2.3 Users’ Reactions During Human-Chatbot Conversations

Talking about chatbots like social actors with a personality and behaviour does not
necessarily imply that computers are human actors. This kind of interaction is a different
and specific world, with its own dynamics and rules.
Several differences have been found between human-human and human-chatbot
communication. Users write shorter messages, have longer conversations and are more
profane with chatbots than with another human. Despite the interaction with an artificial
counterpart, people seem to feel confident and comfortable with chatbots and adapt their
language like in a conversation with children or non-native speakers. The human-chatbot
conversation is often anonymous and this allow a greater use of rude or sexually explicit
dialogue, also because users are curious to test the limits of the virtual agent [17]. During
the design phases is important to try to predict users’ behaviour during the conversation
with a chatbot, in order to prevent this occurrences and to make the whole interaction
more pleasurable.
An interesting aspect emerged during an experiment with the famous chatbot ALICE
is that users anthropomorphised the bot, greeted it, thanked and used other expressions
of courtesy. But, at the same time, users tend to put themselves in a dominant position
and so create an asymmetric relationship as they believe their “natural” intelligence is
superior to the bot’s artificial one [18]. As a consequence they often insulted ALICE as
they felt they can reward or punish the system (e.g. turn off the computer).
The anthropomorphism is a natural tendency of human beings and consists in the
attribution of human features to objects, animals and events. This element it’s also
redeemable in HCI field and can brings to both positive or negative users’ reactions.
Many empirical studies show the similarity between user reactions to humans and arti‐
ficial companions [19].
Humans use stereotypes to simplify their perceptions of others, this could be done
by age, gender or race and could also be positive or negative. This aspects are recog‐
nizable also in conversational agents’ interaction, especially regarding the sex stereo‐
types. For example the attribution of gender is an aspect of the natural human tendency
to anthropomorphize objects. Brahnam evidences the difference between conversations
made with female, male and androgynous chatbot [20]. In all three cases there was
sexually explicit dialogues, but in case of the female virtual agent it happened more
frequently and in more rude ways, pulling out the negative kind of stereotypization [21].
In conclusion, a well-designed chatbot needs a personality and must establish a rela‐
tionship with users, should understand the context and be able to hold a conversation
also about general arguments, not only about its specific domain. It should inspire
108 M. Dibitonto et al.

empathy and at the same time doesn’t try to imitate too much a human (because more
a bot tends to imitates a human and more it is perceived as unnatural and “weird” [22]),
be able to overcome user’s misspellings or other mistakes and diplomatically answers
to potential rude comments. This last instance can be limited by reducing the user’s free
initiative using quick replies instead of free conversation, but on the other side this
restriction risks to make the interaction more similar to a form or fixed menu and not
like a natural conversation [11]. Before starting to design a conversational agent is
important to define the technology, the target, and the typology of chatbot is going to
be realised because there is not only one good solution applicable to all contexts of use.

3 Designing a Chatbot for Universities

3.1 State of the Art

In order to understand the existing scenario about chatbots in university, we have been
analysing the academic literature, paying particular attention to those works focused on
chatbots for Universities. Research shows that a chatbot could innovate campus tech‐
nology, which is often outdated; furthermore it could deliver many benefits to the whole
academic environment, for both students and employees. Nowadays many Universities
provide information and handle frequently asked questions with a physical help desk,
to welcome and assist students and visitors or with a website and a mailbox. Therefore,
in order to manage the amount of daily requests, there must be some staff dedicated to
that purpose [23] and students would need to go to the institute in order to request and
receive real-time and updated information.
Communication media like email can solve the distance issue in some cases and it
is the most frequently used tool to provide information; indeed, email can be an appro‐
priate tool to send a single communication to a large number of students, although it can
be slow and ineffective to manage single requests or to solve specific problems, because
the asynchronicity of the conversation can lead to an unpleasant experience. Therefore
a chatbot can be an useful and effective solution for students, who have the chance to
have immediate and up-to-date information and could ease the staff’s work, which would
be simplified and unloaded.
Despite the restricted academic works, several online sources emphasised that a
chatbot could be an appropriate tool to handle many requests from students.
Chatbots are designed to answer questions or to perform tasks in a specific area. In
the academic environment they usually belong to one of these two categories: the first
one is about chatbots used to help prospective students in finding the best university for
his/her interests; the second category is about chatbots developed for a specific univer‐
sity: some of them act like a virtual assistant for students, while others perform a specific
task. Chatbots, for example, have been used to involve students in Computer Science
topics [24] or to help them in specific courses. These chatbots are task oriented, as the
problem they try to solve are related to specific issues in teaching.
Instead the chatbot we are going to design, LiSA, belongs to the “virtual assistant”
category, whose goal is to provide information and services to students, answering
questions 24 h a day, every day. Information provided by chatbot assistants is usually
 Chatbot in a Campus Environment: Design of LiSA 109

generic, like university’s facilities, admissions, courses. There are different contexts of
use for chatbot assistants in relation to specific moments during the academic year and
specific needs of students. During the application process, a bot could help prospective
students by supporting them during enrollment: the goal to achieve is for students to
retrieve information without any need to browse several web pages looking for
frequently asked questions. As a matter of fact, even though the web is usually a rich
source for this kind of information, and most of universities already provide information
on their websites, sometimes users find challenging to locate relevant information
quickly [25]. Furthermore a chatbot could create a “human-like” connection while
prospective students are seeking for information, establishing a natural interaction.
A chatbot, as a shortcut to get information in an easier, more natural way, is not only
a great advisor to take the first steps into the university environment, but it is also a great
help during the entire time on campus [26]. Many students do not attend lessons daily
and it is important for them to have remotely updated information. Information could
be pull (answer to a certain request) or push, like relevant updates and notifications to
remind students about upcoming deadlines. Pull information usually offered by chatbots
are about university facilities, upcoming events and academic information [27].
Other useful activities are related to push information and notifications. A chatbot
could send notifications to inform students about sudden changes in timetable, for
example, or remember deadlines and important events. The immediacy of the medium
is perfect for rapid information and it could improve the way information is delivered.
Most of chatbots handle pull requests, while only some of them send automatic
messages to students, in order to inform them about important news or sudden changes.
These messages usually concern information addressed to all students, while they do
not provide customized information based on the student’s profile. At the best of our
knowledge we haven’t found chatbots offering transactional services like booking an
exam or paying the annual fee.

3.2 The Design Process for the LiSA Chatbot

Previous works are essential to understand issues and opportunities in designing a
chatbot and to establish next design steps. The focus of this work is to understand the
complex relationships behind a chatbot-human conversation, in order to design a
conversational agent that could be both easy and comfortable to talk with.
Next design steps are focused on observing our target users to understand their needs
and behaviours. To this purpose we designed a questionnaire, useful to understand
students’ need for information, that was submitted as a chatbot in order to observe users
behaviour during the conversation.
In Sect. 4 it will be presented the design process of the chatbot questionnaire, its
aims and the structure we chose to ask questions. Consequently, Sect. 5 will report
questionnaire results, highlighting users’ need for information that will help to define
the database of the LiSA chatbot. Concluding, Sect. 5 will also describe users emerging
behaviours while chatting; this will be essential in order to define chatbot’s personality
and its tone of voice.
110 M. Dibitonto et al.

4 Method

4.1 Survey Bot Design

To explore students’ information needs and their familiarity with chatbot technology a
survey was used. The questionnaire was divided in three main sections. The first section
concerned participants’ personal informations (age, typology of student, etc.), the
second section investigated the informations students are interested to receive or they
looked for, the medium used to get them, and the satisfaction about the service. The goal
of the third section was to probe participants’ familiarity with chatbots and also inves‐
tigate the information they will be interested to receive if the University had one. The
methodology chose to share the survey was LiSA (Link Student Assistant), a Low Level
Dialog System [9] chatbot built with Chatfuel [28] and implemented on the Facebook
Messenger platform (Fig. 1).

Fig. 1. Facebook page of the chatbot LiSA

The chatbot was structured like a survey bot with a System-initiated interaction [11]
and quick replies text bubbles for all the answers, except one. This choice was made to
guarantee an efficient interaction, avoiding misunderstandings due to free conversation,
and for the sequentiality of the questions. The only free answer has been placed at the
end of the survey and concerned which services the participants would be interested in,
if the university had a virtual assistant. The last question was about the satisfaction to
have been interviewed by a chatbot instead of a traditional form.
We created the LiSA’s Facebook page and anticipated in this way the possible future
university virtual assistant to the students. LiSA was provided with a profile picture (a
comic-like character) and a short description of the research in which she is involved.
Because of the young target involved in the research, LiSA used an informal
approach enriched by emoticons, to make the survey more agreeable and coherent with
the mobile messaging applications context and to establish an empathic relationship with
the participants.
 Chatbot in a Campus Environment: Design of LiSA 111

4.2 Participants
In order to investigate students’ information needs and their familiarity with chatbot
technology we interviewed a sample of 100 students. Most of them were Link Campus
University students, while the others were students from different italian universities
and prospective students from high school, who attended some activity at the Link
Campus University. In this way we could investigate the internal students’ information
needs, the orientation informations that would interest a high school student and the
information system offered by other italian universities. The 100 students interviewed
were aged between 15 and 30 years old: 47,4% of interviewed users were aged between
15 and 18 years old, 50,9% aged between 19 and 24 years old and 1,8% aged over 25
years old. Considering that the questionnaire was administered via chatbot on Face‐
book’s platform all the participants were familiar with Facebook Messenger. The
majority of participants were involved to take part to the research personally, during
classes and other activities at the Link Campus University.

5 Results

5.1 User Needs and Habits in Information Request

Most of the students from Link Campus University (88%) claim that they have been
looking for information about University at least once and answered questions about
which tools they have ever used and the level of satisfaction for the answer.

Table 1. Data about the tools students used to communicate with the Link Campus University
Usage Partially satisfied Totally satisfied
Website 93% 28,6% 64,3%
Email 50% 15,4% 77%
Phone 53,3% 7,7% 92,3%
In person 75,9% 26,3% 68,4%

The table above show that students have used different tools to communicate with
the University and that they are quite satisfied with answers they received. A relevant
data comes from phone requests, which revealed the highest satisfaction rate, followed
by “in person” requests, confirming that personal interaction is still preferred by students
to ask for information.
Furthermore the questionnaire highlights that none of the students ever asked for
information on Facebook for several reasons: 41,4% of students say that Facebook is
too informal to communicate with an institution, 41,4% say they do not use Facebook
very often and 17,2% claim they didn’t know they could talk to the University on Face‐
book. Facebook, as a social platform, is considered as an informal communication tool,
not suitable for communicating with University. Relevant data comes from the rela‐
tionship between research content and tool used to search for it, as shown in the table
below.
112 M. Dibitonto et al.

Table 2. Data collected about the way students ask for different kinds of information
Types of information
Enrollment Scholarship Opportunities Events Schedule
Website 21% 40% 36,7% 28,3% 21,7%
Email 18,5% 27,5% 22% 31,6% 27,8%
Phone 33% 12,5% 4,9% 26,6% 29,9%
In person 27,5% 20% 36,4% 13,5% 20,6%

Students were asked how do they usually look for several kind of information. The
table highlights that enrollment generally is the first reason why students actively ask
for information. Students prefer getting information about enrollment in person or by
phone (41%), confirming what other studies found out. Furthermore class schedule is
another issue of primary importance for students, due to the possible sudden changes in
scheduled lessons. Preferred tools to ask for information about class schedule are email
and phone, because allow students to get information without the need to go to the
university.
Further questions focus on the information Link Campus University already offers
to students and the way they deliver it. In this case we will consider only answers from
current students of Link Campus University.
Most of the students claim that Link Campus University offers an information and
notification service (81,8%) that reminds about important events and deadlines. They
say that information is delivered via email (70,8%), phone (7,7%) or both (21,5%). The
majority of the students are very satisfied from this communication service (69,2%),
while 26,9% claim they are not very satisfied and 3,8% are not satisfied at all. Email
seems to be the most widely used tool to provide personalised information at Link
Campus University: 64% of the students say they receive personalised information and
36% say they do not. These students were asked further questions about their opinion
about customised information and 66,3% of them say they would find useful receiving
personalised information from the University.
Both current and prospective students answered questions about their knowledge on
chatbots. Only 51,9% of students claim they know what a chatbot is, while the others
have never heard about it. It is interesting to compare answers basing on age: most of
younger users from high school know what a chatbot is (70%) while only the 56,3% of
university students know it. This data would suggest that younger users are more sensi‐
tive and reactive to new communication tools.
A part of those who know what a chatbot is, claim that they have been using it to
contact some customer service (65,4%). Students who do not know anything about
chatbots say they have never used it because they use telephone to contact customer
services (62,9%) or email (22,9%). 14,3% of users say they believe that chat is not an
appropriate tool to contact a customer service so they have never used it.
Users who used chat tell they have talked with humans (38,5%) and chatbots (39,5%)
and they could distinguish between human conversation and chatbot conversation
because it was specified (70%), or the chatbot didn’t understand requests (10%) or it was
repetitive (20%). It is important to highlight that most of the users who had a conversation
 Chatbot in a Campus Environment: Design of LiSA 113

with a chatbot evaluate it as satisfying (50%) or indifferent (40%), while only 10% define
it as frustrating.
All the students interviewed have listed the information they would like to receive
by the University via chatbot; answers were submitted in natural language, so data was
collected and aggregated, in order to define macro-categories of information (Fig. 2).
In conclusion this questionnaire is a valuable contribution in finding out students’
need, desires and current lacks in information delivering and using a chatbot was useful
to directly observe students’ perception and behaviour with this communication tool.

Fig. 2. Data about what type of information students would receive by an university chatbot

5.2 User Behaviour During the Conversation

Previous works show that people have different behaviours during an human-chatbot
communication than during human-human communication and the analysis of the
conversations with LiSA prove that. Most students correctly answered the questions
using quick replies, often thanked LiSA and used other expressions of courtesy, enriched
by smiling emoticons. This shows a tendency to anthropomorphize the chatbot, even if
it has a low level of complexity. Other participants, instead, abandoned the conversation
when LiSA asked to write personal opinions freely, showing an opposite, unconcerned
reaction.
At the last question about rating the experience with LiSA, the 59,6% of participants
valued it like interesting, 29,8% enjoyable and 6,4% indifferent. A marginal number of
students rated it as annoying and boring; these opinions are closely linked to situations
when LiSA looped or got wrong, making users frustrated or to the participants’ curiosity
about the chatbot response. Aggressive reactions were registered as well: some students
(6%) insulted and offended LiSA, flirted with her or talked about physical appreciation
despite LiSA has a simple and not sensual graphic appearance. In a first phase we have
not included filters concerning rude or sexually explicit words, because the students
were mostly involved personally in a university context and because their Facebook
114 M. Dibitonto et al.

profiles were not anonymous. After the first impolite answer we decided to provide LiSA
with a sensitive personality, hurted by aggressive behaviours. This personality change
led to different reactions: some students apologised with LiSA, while others tried to
excuse their own behaviour with the curiosity to testing the limits of the chatbot. Rude
words were used by both female and male, especially when LiSA looped or got wrong,
as an evidence of dominant position. Flirting was only observed in men, confirming
Brahnam [20] gender stereotypes observations.
Students who showed aggressive behaviours were especially younger prospective
students, while Link Campus University students generally had a polite behaviour and
completed all the questionnaire.

6 Discussion

Testing a chatbot gave us many interesting sparks in terms of information needs and of
personality design guidelines.
The information that students are used to ask for, described in Table 1, are essential
to define the information database that the LiSA chatbot will necessarily include.
Table 2 will be an interesting hint to design further services that LiSA could provide.
For example, it was observed that students would find useful to have a chatbot sending
information and notifications about lessons, especially about changes in class schedule.
This kind of information should be delivered in push mode, in order to immediately
advise students.
Furthermore, students would like to be advised about University’s events and activ‐
ities. Also this information might be delivered in push mode and it could be personalised
on student’s interests and academic career. General information as well as information
about enrollment will be included as well, but they might be available on request,
supported by push notification about important deadlines and messages, already deliv‐
ered by email.
Lastly, we will take into account further useful services, like exam booking, currently
available on the student’s personal page on the University online platform.
Analysing conversations allowed us to observe recurring behaviours, like sexually
explicit and rude reactions, already proved in previous works. According to the evidence
it is necessary to properly design LiSA’s personality, enhancing empathy and sensitivity.
Conversation should be well balanced and polite as an human-human conversation and
this goal can be reached only by designing with empathy.
It can be assumed that some behaviours occurred because students believed they
were anonymous, despite the conversation was actually a questionnaire presented by an
University and above all conducted on Facebook, where identity is perfectly recognis‐
able. This data highlights that some students improperly used the chatbot tool and that
they did not care about their privacy and identity, especially when they were frustrated
by errors or loops during the conversation.
 Chatbot in a Campus Environment: Design of LiSA 115

7 Future Works

The next steps in designing the University virtual assistant will be based on the collected
data analysis and other research studies cited in this work. First of all we will decide the
gender and visual look of the chatbot and consequently the personality and tone of voice
used during the conversations with students (formal or informal, use of emoticons, etc.).
We will test female, male or neutral gender for the virtual assistant, in order to observe if
there will be substantial differences in sexually explicit and rude reactions; then we will
shape the final chatbot aspect and personality. The virtual assistant will be provided with
an own biography and the capacity to answer general knowledge questions besides about
specific university’s topics. Personality is essential to shape an empathic relationship with
the user and we would like that LiSA become a useful personal assistant for every Link
Campus University student, able to give real time personalized information in a pleasant and
informal way. We will implement filters and emphatic answers to make LiSA able to repair
potential dialogue mistakes and to diplomatically answer to rude or sexual provocations.
We would decide the right proportion between quick replies and free dialogue
according to the type of information to give; then we will define whether if the system
or the user will lead the conversation and in which cases.
At the end of the design phase we will realise the first prototype of LiSA for testing
it with a sample of students, in order to understand its effectiveness and thoughtfulness
before the final release.

References

1. Mastercard Developers. https://developer.mastercard.com/product/masterpass-chatbot.

Accessed 10 Jan 2018
2. De Angeli, A., Lynch, P., Johnson, G.: Personifying the e-market: a framework for social
agents. In: Hirose, M. (ed.) Proceedings of Interact 2001, pp. 198–205. IOS Press (2001)
3. Weizenbaum, J.: ELIZA - a computer program for the study of natural language
communication between man and machine. Commun. ACM 9(1), 36–45 (1966)
4. Wallace, R.S.: The anatomy of A.L.I.C.E. In: Epstein, R., Roberts, G., Beber, G. (eds.) Parsing
the Turing Test, pp. 181–210. Springer, Dordrecht (2009). https://doi.org/
10.1007/978-1-4020-6710-5_13
5. Shawar, B.A., Atwell, E.: Chatbots: are they really useful? LDV Forum 22(1), 29–49 (2007)
6. Følstad, A., Brandtzæg, P.B.: Chatbots and the new world of HCI. Interactions 24(4), 38–42
(2017)
7. Xu, A., Liu, Z., Guo, Y., Sinha, V., Akkiraju, R.: A new chatbot for customer service on
social media. In: Proceedings of the ACM Conference on Human Factors in Computing
Systems. ACM (2017)
8. Brandtzaeg, P.B., Følstad, A.: Why people use Chatbots. In: Kompatsiaris, I., Cave, J.,
Satsiou, A., Carle, G., Passani, A., Kontopoulos, E., Diplaris, S., McMillan, D. (eds.) INSCI
2017. LNCS, vol. 10673, pp. 377–392. Springer, Cham (2017). https://doi.org/
10.1007/978-3-319-70284-1_30
9. Schumaker, R.P., Ginsburg, M., Chen, H., Liu, Y.: An evaluation of the chat and knowledge
delivery components of a low-level dialog system: the AZ-ALICE experiment. Decis. Support
Syst. 42(4), 2236–2246 (2007)
116 M. Dibitonto et al.

10. Meisel, W.: Specialized Digital Assistants and Bots: Vendor Guide and Market Study. TMA
Associates, Tarzana (2016)
11. Duijst, D.: Can we Improve the User Experience of Chatbots with Personalisation? (2017)
12. Dale, R.: The return of the Chatbots. Nat. Lang. Eng. 22(5), 811–817 (2016)
13. Fast Company Co.: Design. https://www.fastcodesign.com/3048139/what-is-zero-ui-and-
why-is-it-crucial-to-the-future-of-design. Accessed 9 Jan 2018
14. Sansonnet, J.P., Leray, D., Martin, J.C.: Architecture of a framework for generic assisting
conversational agents. In: Gratch, J., Young, M., Aylett, R., Ballin, D., Olivier, P. (eds.) IVA
2006. LNCS (LNAI), vol. 4133, pp. 145–156. Springer, Heidelberg (2006). https://doi.org/
10.1007/11821830_12
15. Kuligowska, K.: Commercial Chatbot: performance evaluation, usability metrics and quality
standards of embodied conversational agents (2015)
16. Holtgraves, T.: Communication in context: effects of speaker status on the comprehension of
indirect requests. J. Exper. Psychol. Learn. Mem. Cogn. 20, 1205–1218 (1994)
17. Hill, J., Ford, W.R., Farrera, I.G.: Real conversations with artificial intelligence: a comparison
between human-human online conversations and human-chatbot. Comput. Hum. Behav. 49,
245–250 (2015)
18. De Angeli, A.J., Graham, I., Coventry, L.: The unfriendly user: exploring social reactions to
Chatterbots. In: Proceedings of the International Conference on Affective Human Factors
Design. Asean Academic Press. Han, S. K., YounGi, London (2001)
19. Rickenberg, R., Reeves, B.: The effects of animated characters on anxiety, task performance,
and evaluations of user interfaces. In: Letters of CHI 2000, pp. 49–56 (2000)
20. Brahnam, S.: Gendered bods and bot abuse. In: CHI 2006 Workshop Misuse and Abuse of
Interactive Technologies, Montréal, Québec (2006)
21. De Angeli, A., Brahnam, S.: Sex stereotypes and conversational agents. In: Proceedings of
Gender and Interaction: Real and Virtual Women in a Male World, Venice, Italy (2006)
22. Ciechanowski, L., Przegalinska, A., Magnuski, M., Gloor, P.: In the shades of the uncanny
valley: An experimental study of human–chatbot interaction. Future Gener. Comput. Syst.
(2018)
23. Memon, A.R., Chowdhry, B.S., Shah, S.M.S., Memon, T.R., Shah, S.M.A.: An electronic
information desk system for information dissemination in educational institutions. In: 2nd
International Conference on Computing for Sustainable Global Development (INDIACom
2015), pp. 1275–1280. IEEE (2015)
24. Benotti, L., Martínez, M.C., Schapachnik, F.: Engaging high school students using Chatbots.
In: Proceedings of the 2014 Conference on Innovation & Technology in Computer Science
Education, ITiCSE 2014, pp. 63–68. ACM (2014)
25. Ghose, S., Barua, J.J.: Toward the implementation of a topic specific dialogue based natural
language Chatbot as an undergraduate advisor. In: International Conference on Informatics,
Electronics & Vision (ICIEV) 2013, pp. 1–5. IIIE (2013)
26. ECampus News. https://www.ecampusnews.com/campus-administration/chatbots-campus-
students/. Accessed 1 Feb 2018
27. Ranoliya, B.R., Raghuwanshi, N., Singh, S.: Chatbot for university related FAQs. In: 2017
International Conference on Advances in Computing, Communications and Informatics
(ICACCI), Udupi (2017)
28. Chatfuel. https://chatfuel.com/. Accessed 15 Feb 2018