1.
TaskBot: Understanding
Chatbot-mediated Task
Management
Carlos Toxtli, HCI Lab - West Virginia University
Justin Cranshaw, Microsoft Research
Andrés Monroy-Hernández, Microsoft Research*

2.
Example
Task Management over group communication
channels
● Where: Microsoft Teams
● What: Alice is asking Bob to complete and send
his Tax Form
● True story ...

3.
Reflection
Social friction, emotional burden, etc.
And it was a 1:1 conversation ...

4.
Task Management
● Teams might discuss a project over a group channel, decide who gets to
work on what, then add those tasks to a task management system.
● These interruptions can add up to reduced productivity and increased
stress in the workplace (Czerwinski 2000, Iqbal 2007, Mark 2008)
● Because of the friction of switching contexts, tasks that need to be tracked
end up getting lost or forgotten because people don’t switch contexts to
track it.
● Extra work of tracking tasks requires specialized role (e.g. PM), which
often does not exist in small teams.

5.
Task Management Solutions
● Traditional task management tools:
● Reduced friction by enabling adding tasks anywhere, mobile:
...
...

6.
Is there any other way to do it?
This is why we proposed TaskBot, let me explain what it does and how it
works by an example.
Same case, Alice is asking Bob to complete his Tax Form but this time through
TaskBot

7.
Reflection
It reduced the number of Alice´s interactions
Works for teams with multiple members

8.
Why Chatbots? Tools VS Assistants
(Cranshaw 2017)
Tools are
● Frictionful, requiring interruptions and context switches
● Inflexible, tailored to a specific set of needs
Assistants are
● Frictionless: in situ delegation doesn’t require interruption
● Flexible, able to respond to rich preferences and workflows

9.
Implementation
No specific rules were given
to users. We Iterated over
Natural Language Processing
models by training an existing
service
Interfaced with an existing
task management tool. No
previous setup from the
users is required
Chatbot
infrastructure

10.
User Study’s Goal:
Understand how bots can mediate task management through
deployment of TaskBot in real working teams.

11.
User Study’s details
● Length of study: 1 week
● Teams:
○ Number of teams: 8
○ Teams were asked to create at least 3 tasks over the course one week
○ Team sizes from 2 to 5 people, with 19 people in total
○ 5 of the 8 teams were hierarchical
● Types of workers: manager, subordinate
● Surveys (pre and post)
● $20 gift card

12.
Usage summary
Teams 1 2 3 4 5 6 7 8
Members 3 2 3 3 2 2 3 5
Tasks 20 5 12 9 20 6 5 12
Messages 142 70 139 131 267 54 86 170
88 tasks created (4.4 per user on average)
65 tasks marked as “completed”/
TaskBot received 177 messages from participants
○ 22% of these messages were tasks assignments
○ 78% transactional communication.

13.
Usage by individuals
● On average
○ Users sent 12 messages to TaskBot, while users received 54 messages from
TaskBot
○ Users received 16 reminders to complete their assigned tasks.
6
4
11
Tasks

14.
User feedback
● 6 of the 11 people rated the felt more productive,
“I felt more productive because the conversation side felt much easier and
made me less stressed with getting things done.”
● 8 people will use it in the future, however
“I liked how easy it was to assign a task.”
● 4 people reported finding it annoying
“I did not like that it told me all my pending tasks. I got annoyed with just 4
at one point.”

15.
User feedback
● Main features perceived by users
Reminding other people about tasks (10 of 11 people)
Tracking the progress of tasks (9 of 11 people)

16.
Patterns observed
1. Handling human-like interactions with bots
2. Supporting self-communication
3. Hierarchical task-assignment
4. Failing gracefully
5. Dealing with human ambiguity
6. Identifying people’s name in conversations
7. Handling multi-threaded conversations

17.
Supporting self-communication
Even though the user training focused
on how to assign tasks to others, five
users asked TaskBot to create
reminders for themselves:
Designers of social chatbots should
assume that bots would also be used
for self-communication, either as a
way to test the system or as a
practical use of the tool.
“Remind me at 10:15 to leave”

18.
Resolving name ambiguity
People didn’t always use the special
@mention syntax and this created
problems for TaskBot. 40% of the users
forgot to type the @ before the name.
Designers should find ways of nudging
users to mention people in the ways
communication channels expect (e.g.
using the at sign), or create smarter
ways of detecting when a person
might be mentioned in a message.
“Hey John, can you finish your tutorial?
cc @TaskBot”
Who?

19.
Handling multi-threaded conversations
The biggest challenges for TaskBot and
other bots is the difficulty of
maintaining multiple active
conversations at the same time.
Designers should invest in technology
for determining which active
conversation thread a new incoming
message belongs to
“I’m done with the task,”
Which task?

20.
Organization hierarchy
● People assigned tasks across
different hierarchical levels.
○ 35% of tasks that were assigned to
managers by their subordinates
○ 31% assigned to subordinates by their
managers
○ 34% were among people at the same
level.
● Managers and subordinates used
TaskBot differently
○ 83% of requests that were upward in
hierarchy (from subordinate to
manager) were reminders
○ 47% of tasks that were downward in
hierarchy (from manager to
subordinate) were reminders the
rest were direct assignations
Designers of social chatbots should expect different uses of the same bot based
on people’s hierarchy.

21.
Conclusions
Deployed TaskBot to understand how bots can mediate task management.
Ran a study with 8 real teams working on real projects.
Defined guidelines from users’ interactions.
Future work should be done to further explore hierarchical interactions and
dynamics over different channels and other tools integration.

22.
Thank you!
Carlos Toxtli-Hernández
HCI Lab - West Virginia University
http://www.carlostoxtli.com
carlos.toxtli@mail.wvu.edu
@ctoxtli