Chat of machine language, natural language processing and human

Chat bots in HR

and Efficiency*, Yay or Nay?

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!

order now

Maryum Arshad

(Research Skill Seminar)



[email protected]



Abstract- Artificial Intelligence is one of the
emerging fields and most widely worked-on in recent decades. It is a
combination of machine language, natural language processing and human computer
interaction, mimicking the functionality of a human brain to perform different
tasks.  This paper is being written to
discuss about the practical uses of “Chatbots in HR”, its limitations and
drawbacks and the amount of work and technology efficiency achieved through
this system.
Keywords: Artificial
Intelligence, chatbots, HR, Intelligent Systems, Automated System, machine learning,
natural language processing.

Introduction to

intelligently designed robots (hardware/software) working on the principle of
self-learning with the help of already given information/knowledge and
predicting the future events fall under the category of Artificial
Intelligence. Chatbots are one such example of software based artificial agent.
These are so widespread and intelligent that you’ll have no idea to whom you are
chatting or talking while interacting over internet. With the excessive
dependence on technology, people find it easier to rely on automated systems
rather than doing tasks manually. From shopping to marketing/ publicizing one’s
products, chatbots are found everywhere. Have you ever wondered that how do you
get suggestions while shopping online or while searching anything over the
internet- whether it’s a movie, university or a restaurant and then the next
time you connect to internet your web browser/websites/apps are full of those
similar advertisements, how does it happen? The chatbots spy on your behavior
and recommend you the similar products to develop your interest in their
sites/products/offers. The deployment of chatbots in those places is not only
economic, but time saving and interesting as well. Chatbot programs are fed
with many different natural languages so that they can interact and learn from
different kinds of people from different parts of the world. They not only work
using already given information but also do run time functions like calculating

expression or setting a reminder etc. There are many popular recent examples of
successful chatbot (which are discussed later in this paper). Artificially
designed such systems may fail e.g. Microsoft’s chatbot named as ‘Tay’ (acronym
for ‘Thinking about You’) that was created on twitter to interact with people
and to learn their behavior,  but the
Microsoft had to took it down in less than 24 hours because it was attacked by
some trolls who made Tay to learn abusing behavior. They re-released it with
some privacy added to it after few months but still couldn’t work as expected
and ultimately they had to take it down. So it clearly means that artificial
agents need a lot to be fed to them in their knowledge base to avoid such
circumstances. IBM is also proving an open source to design your own customized
chatbots and you don’t even have to have prior knowledge of machine learning, and
you can use those chatbots in different websites of yours like online shopping
and others. Such intelligent systems are economic too e.g. once a chatbot is
made and implemented, the cost required to add additional features will be

Chatbots in HR


Coming to the main discussion, usage of chatbots
in HR -whether these are useful or not, or are being misused- I’ll like to
start with discussing how such a system works, followed by its functionality, examples
and drawbacks, future of such a system, what actually HR professionals think
about it and whether candidates are comfortable being interviewed by such an HR
representative or not.


A: Functionality:

Chatbots are no doubt one of the most
well-known living examples of Human Computer Interaction (HCI). I have studied
about a chatbot named SARANG, which is based on Artificial Intelligence Markup
Language (AIML) that is originally designed to develop automated chatbots and
provides many open source functionalities. The purpose of AI is to define work
into simple differentiated modules. The system works successfully due to mutual
interaction of these modules. The logical/functional division of this program is
as follows:


Welcome Screen:

is the introductory screen that asked the used his/her introduction and a short
biodata like name or field on profession for further ‘intelligent’ discussions.
That’s how the name of user is stored in data file for later use.


Chat Interface:

is the screen where texts with a user are exchanged. This interface is fed with
a manager that controls reply on the basis of data stored in its knowledge
base. It works on the strategy of ‘search and display’ on the basis on rules
stored in it.


The Database:

is like a ‘memory’ to the chatbot and does the same work as human brain. It has
stored in itself answers to many different possible questions that can come
from user end and the capability to answer such questions. There are two main
things to consider: to avoid the redundancy i.e. giving the same answer again
and again or giving the same answer to different questions and to make logic
between more than one answer during a conversation i.e. semantics has to be
built in replies. Dealing with the information about the employees contains a
lot of data, and obviously efficient systems are required to interpret it too.


Intelligent Applications:

additional productive applications are added to the AI agent like a simple
calculator/calendar/dictionary or games to make it more interesting. So that if
user comes up with a question that includes date/ time or their calculation
then this agent is able to answer intelligently if it is fed with such additional
functionalities. Not only this, due to courtesy of these agents you can even
set your alarm or reminders, create notes or memos etc. and can even pop it up
on your gadget (e.g. computer, mobile or tablet) display with your permission.
Doesn’t that sound great to open your laptop and getting a pop up which says
‘Your salary has been credited to your account’, certainly it does.


Handling the

a user puts a tricky question or a cross question to which the answer is not
stored in database, this intelligent system handles with its smartness and put
an illustration that user couldn’t expect, hence the user is satisfied even
with a smart question.


Data handling:

such example is no one wants to tell you his/her name every time you meet. Same
is the case here; a user wants to be remembered so that he/she doesn’t have to
introduce to the chatbot again and again. In such a case data handling is
required, with the help of this efficient feature the chatbot is able to
remember the name and other relevant information so the next time when this
user appears, the chatbot won’t have to waste time in introduction and they can
start with a real conversation. This also helps to modify the knowledge base to
answer that particular user specifically. Hence the smartness and intelligence
is ensured. One such example in HR is handling the data of an employee with all
the previous records so whenever any information is inquired by a manager or by
employee himself, the chatbot will give the information about the concerned
person other than giving the irrelevant data.


Dealing with the

of the worst case examples that could lead to an error is diving any number
with zero. To calculate such a query the chatbot will be lost in indefinite
divisions. So the system should be smart enough to deal with such a case as
well. In case of HR the error can be checking the record or personal
information of any other employee thus violating the rules or standards of the


Diagrammatic Structure:




Fig. 1 Diagrammatic approach to functionality
behavior of a chatbot


B: Working an HR-chatbot

There are two kinds of chatbots
depending upon the type of implementation- first ones are given structured questions
and answers to work on and the second ones are based on machine learning. Prior
has the limited knowledge base and skill set and can response correctly only to
specific input quests whereas the latter is the example of artificial
intelligence agent, which not only works on the basis of given knowledge base data
but also learns new behavior and thinks like a human while taking a decision. It
also has the ability to store new behavior in its knowledge base skill set once
it has learned during a real interaction with people. In either case, the smartness
and efficiency of a chatbot depend upon the smartness of


As the chatbot can converse with you in
text, voice or video, it is believed that soon they will be able to read human
emotions as well and will give the results with better understanding. It tests
the person by asking different kinds of questions and determines the field in
which the person is more influenced.  Implementing chatbot in HR serves many
purposes e.g. if a person is ill and he informs to company about leave via an
application particularly designed for employees, and he says that he is not
able to join office today because he is ill, he’ll get a reply like ‘It’s sad
to hear that’ or ‘Get well soon!’ and at the same time (after your
unavailability has been confirmed) your absence will be sent to respective
manager and he’ll assign you some other time. Certainly no HR person is
duty-less enough to have such a conversation so this means that all these tasks
i.e. conversation and reporting are done by chatbot, which is faster, simpler,
effortless, trouble-free and economic. In addition to this task, an HR chatbot
can receive audio messages, interpret it and forwards to the concerned
person/department and can send an email as well in this regard. These robots are
fed with AI along with NLP and optimized search so thus are also able to answer
certain questions about an employee like salary, benefits and other HR related
things. Furthermore, unlike humans, chatbots are capable to deal with ‘any’
kind of employee even if the person has repeated a question many a times about
a certain problem within a department.


For Recruitment phase, chatbots are able
to filter resumes by their ability to read, understand and interpret text,
followed by giving the list of selected candidates who are then called for the
interview e.g. Mya. For
interviewing, chatbots are fed to ask several questions for initial recruitment
like educational background, desired location of work, tools and technology
with hands-on experience, preferred hours of working etc., records the answers
and forwards results to higher authorities for further processing of successful
candidates and discards the below standard data. This saves the time of HR team
for initial filtering of candidates. After an employee has joined a firm,
chatbots are able to tell them about onboarding information e.g. history, how
and where to go for a certain piece of work.

Deploying chatbots in HR serves other
purpose too like 24/7 availability and faster access of information, customized
information for each person, handling FAQs in a better way and foreseeing a
potential issue based on data analysis with in any department/organization.


C: HR’s point of view

It is considered as useful to implement
chatbots instead of human HR team because chatbots are time saving and can work
faster than people. The main issue is to make a link between HR enterprise
software and customized chatbot i.e. to integrate them together and maintaining
their mutual operation without any error or misbehaving is the actual task,
although chatbots are efficient enough to provide generic information even if
they are not customized.


D: Employee’s point of view:

HR chatbot provides at the moment and
friendly information to the employees. Moreover chatbots are irrespective to
judge an employee’s personal choice on the basis on questions asked, that’s why
some employees prefer it as it saves them from embarrassment maintaining their
dignity and reputation.


E: Candidate’s point of view:

A candidate appearing for an interview
may not feel comfortable being interviewed by a machine. Sometimes he may get
confuse during answering phase and need a time to think about it but what if
the chatbot is fed with limited time for each question and he may come under
pressure in such a scenario. Maybe the candidate is capable enough for the vacant
seat but the situation isn’t in his favor, so in such a case it is a loss for
both of the candidate and the company.



ELIZA: It is a
chatbot that was first documented by Joseph Weizenbaum. It is the simulation of
a psychotherapist to conduct a psychiatric interview. It takes textual user
input and analyzes it using its intelligence.

ALICE: It is the first
AIML based artificial agent. It was just built to show the ability of
‘supervised learning’ and proving that it is capable of more appropriate

Ask Ivy: It is a virtual
HR agent by Intel that works on the principle of Ai, natural language
processing and optimized search. It is local software that works on their
intranet to reply to employees’ queries regarding pay, benefits and other HR
programs etc.

Mila: It is the’s chatbot to track the sick leave of employees and rescheduling
and reporting about it.

Mya: It is a
recruitment chatbot that shortlists the applications based on resume’s textual
screening, manages tests, conducts interviews and also able to give tips to

Sergeant Star (Strongly
Trained And Ready):
It is used in recruitment in American army.

Meekan: This chatbot
provides you the facility to schedule a meeting or an event between different
employees. The plus point is that it understands the natural language requests
made my human users.



During interview we can see the
expression of an interviewer and can change our response but a robot lacks such
ability thus there will be no 2nd chance for the candidate. It’s
like the ‘Now or Never’ scenario, if you lose your chance, neither you can
change your answer at the moment nor you can give additional answer once your
time is gone.

Secondly, maintaining the integrated
components and updating it requires a technical team, if you are saving your
finance by downsizing the HR team, you are paying for the technical team, and
if you don’t have any such permanent team, you’ll need an agreement to a
technical team/software house who’ll be responsible for maintaining it, which
surely costs less than having a permanent team 24/7. The only hazard is that
you’ll have to look for recognized company e.g. what if you get your software
and after few months the company dissolves? This is the trouble that needs to
be sorted out in preplanning.

Furthermore, working on data security to
increase the secrecy of user’s data, is another main concern in this regard.


Improvement and Future

Here’s an interesting example referring
to the future of chatbots. Consider a scenario in which a candidate is being
interviewed by a chatbot for a particular post in a company. The company has an
intelligent chatbot fed with all relevant questions to interview the candidate,
and in a normal working condition, chatbot will be questioning and a candidate
will be answering, but what if the candidate himself has a chatbot to answer
the questions. How a chatbot will be able to check if there is a person on the
other side or chatbot as well?  There are
few techniques in practice on different websites like ‘prove that you are not a
robot’ followed by examples of simple mathematical questions or images. A
chatbot vs chatbot conversation could either be timewasting (giving no result) or
exceptionally perfect (technology vs technology), but in either case, it would
be devastating and useless (in my opinion). It is the developer who is going to
decide about adding new features so always look for the possibilities to make
it better.


Results & Discussion

Considering all the favors and odds of
the chatbots in HR, I’ve come to the conclusion that it is convenient or user
friendly for employees to use, efficient for the HR department and economic for
the owner of the company. But this is not the end. AI has spread everywhere,
from a simple shopping website to a complex HR chatbot; all the automated
processes are handled with the help of AI. In a near future, I can see the
world being taken over by automated technology and this is all the effect of artificial


In a conclusion, I would say that the benefits being obtained by these smart
systems overcome their drawbacks. With the extensive work in this field the
engineers are quite sure to make it more reliable, secure, smart and
intelligent. They are economic as
major cost is only at the initial stage while developing them and the
maintenance cost is negligible comparing to the development cost. Furthermore
you do not need to develop such a system again and again from a scratch, once
you have a working system you can add modules whenever you want. Many open
sources chatbots are available; you can just customize it according to your own
needs and requirements and use it. The chatbots
are also timesaving; the human HR
can even spend many days to shortlist hundreds of candidates in initial
screening whereas the chatbots are able to do that at the moment. Not only
this, human HR will first spend time on interviewing then on analyzing the
reports obtained from interview, whereas the chatbots are able to do both tasks
at the same time, i.e. by the time the interviews are over, you can get the
result list of better candidates as well. They are also efficient and convenient
to use.


1. Kong, Yean Hwei, Amy Kona, ‘Chatbots
in HR: Improving the HR Experience’, January 2017. Online Available:

2. Tom Haak, ‘The Invasion of
Chatbots’ (HR Trrends 2017, 16), June 2017. Online Available:

3. Matt Chessan, ‘the MADCOM Future;
Hpw Artificial Intelligence Will Enhance Computer Propaganda, Reprogram Human
Culture, and Threaten Democracy…and What Can be Done About It’, September 2017.
Online Available:

4. Onlim, ‘How do chatbots Work’,
Online Available: