Seamless Group Trip Planning: Connecting Personal Inspirations to
Unified Decisions

by
Dhanvi Patel
M.Des (Interaction), Emily Carr University of Art + Design, 2022-2024

A CRITICAL AND PROCESS DOCUMENTATION THESIS PAPER SUBMITTED IN PARTIAL
FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

MASTER OF DESIGN

EMILY CARR UNIVERSITY OF ART + DESIGN
2024

©Dhanvi Patel, 2024

Abstract
This study aims to conceptualise a digitally collaborative trip planning experience for
travellers seeking personalised recommendations and assistance in group decision-making
before and while travelling. The proposed outcome uses AI-generated preferences
gathered from their selected social media posts to make personalised recommendation
and an AI-chatbot to facilitate group trip planning.

To achieve this objective, key users were defined as tech-savvy, Gen Z individuals who
prefer group trips and their behaviour and preferences were studied through literature
review, participatory research, and auto-ethnographic observations. The key issues
identified for the defined users before and during their trip included finding personalised
recommendations and decision-making in a group while being remotely connected.

Based on the insights of this research one of the outcomes is an interactive prototype of
the key features that are potential solutions for relevant recommendations with AI and an
example case of AI chat-bot facilitation for ease of navigating the group planning process.
This prototype includes the following key features:
1. Integration of saved social media posts to create a personalised itinerary, serving as
a starting point. Furthermore, same data can serve as a resource for AI-chatbot to
provide relevant suggestions.
2. An AI-chatbot that facilitates group discussion for encouraging diverse opinions,
even participation and unbiased conflict resolutions.
3. Ability to modify itinerary by adding options and facilitating voting with a map view.

Travel-related service providers can benefit from insights gained through this study,
enabling them to better understand and engage with the users. Furthermore, the design

2

process and interface design decisions can serve as a valuable reference for anyone in the
field of user-centric application development which involves AI.

Acknowledgements
This thesis would not have been possible without the support and participation of
numerous people.

Firstly, I would like to extend my sincerest gratitude to my supervisor Haig Armen, whose
unwavering support and guidance have been instrumental in shaping this thesis. Your
wisdom and expertise have not only inspired me but also paved the way for my exploration
and understanding of complex design principles.

I also want to express my sincere gratitude to all the participants for giving their precious
time and insightful feedback. This thesis would not have been feasible without your
eagerness to participate and your passion for the travel and technology. Your viewpoints
and experiences have served as the foundation for my research.

Moreover, I would like to thank my internal supervisor Eugenia Bertulis, whose insightful
inputs and constructive criticism have been crucial in filling the missing gaps in my
research.

Finally, I would like to acknowledge my cohort and peers for their thoughtful feedback on
my work during multiple review sessions.

Thank you all for playing a pivotal role in my academic journey.

3

Table of Contents
Abstract

2

1. Introduction

8

1.1 Project Rationale

8

1.2 Research Questions

9

1.3 Objectives

9

1.4 Scope and Limitations

10

1.5 Structure & Methodology

12

2. Setting the Context (Secondary Research)

13

2.1 Demographic Behaviour & Preferences for Trip Planning

13

2.2 Experiences Enhanced by Artificial Intelligence

16

2.3 The Balance Between Pre-planning and Spontaneity

19

2.4 Remote Collaborative Planning

20

2.5 Comparative Analysis

21

3: Primary Research

26

3.1 Informal Interviews

26

3.2 Experimenting with ChatGPT

27

3.3 Participatory Research

30

3.4 Auto-ethnographic Observations

35

4. Synthesis

37

4.1 Insights

37

4.2 Persona

39

4.3 User Journey Map & Scenario

42

5. Outcome

46

5.1 Ideation

46

5.2 Wireframes

49
4

5.3 Final Concept

51

5.4 The Working Explained

56

5.5 Testing and Iterating

58

6. Reflections and Future Directions

60

6.1 Reflection on the process

60

6.2 Designers and AI

60

6.3 Future Directions

61

6.4 Ethical Considerations

62

7. Conclusion

64

References

66

Appendices

70

5

List of Figures
Figure 1 Venn diagram representing objectives.

10

Figure 2 Data visualisation of travel frequency & motivations.

15

Figure 3 A conceptual approach to travellers planned and unplanned attraction visits
(adapted from Hwang, 2011, p 289)

19

Figure 4 Smart search with AI Assistant

22

Figure 5 Group collaboration with voting.

22

Figure 6 Resources from social networks and blogs.

23

Figure 7 Search with AI

23

Figure 8 Collaborative Itinerary

23

Figure 9 AI Assistant which in not context aware.

24

Figure 10 Poll feature for group decision making.

24

Figure 11 Mood board style itinerary.

24

Figure 12 Budget Calculations

28

Figure 13 List of activities.

28

Figure 14 Detailed itinerary.

28

Figure 15 Itinerary based on preferences.

29

Figure 16 Result of activity with 2 participants.

31

Figure 17 Result of activity with 1 participant.

31

Figure 18 Result of activity with 3 participants.

31

Figure 19 Definitely planned and flexible areas of the trip represented with post it.

39

Figure 20 Persona 1: Amanda, The Instagram-Savvy Scenic Lover

40

Figure 21Persona 2: Rachel, The Budget-Conscious Explorer

41

Figure 22 Journey Map of Amanda and Rachel.

43

Figure 23 A Storyboard of AI chatbot facilitation through an example.

45

Figure 24 Ideation sketch of a chat-based approach.

48

Figure 25 Adding options for places initial attempt.

49

Figure 26 Adding options for places revised for a better experience.

50
6

Figure 27 Feature considerations for a chat-based solution: AI-facilitation for choosing
accommodation.

51

Figure 28 prototype design representing flow for relevant itinerary generation.

52

Figure 29 Features for ease of collaborative planning.

54

Figure 30 Screens showing AI-chatbot facilitating group discussion for deciding
accommodation.

55

Figure 31 System architecture diagram.

57

Figure 32 iteration screens of collaborative panning and AI collaboration.

58

7

1. Introduction
“What we’re imagining in a future with gen AI or AI in general is that you start with something
much more free-form and say, for example, “I’m looking to plan a trip with my family to New
Orleans for a week in October. Can you help me find a hotel that has a pool for my seven-yearold and is within walking distance of the French Quarter?”
Wouldn’t that experience be much easier in terms of trying to figure out where you want to stay
and what you want to do, as opposed to getting a list of a thousand hotels in an order that may
or may not meet your specific preferences and what you want out of that trip? “
- Vik Krishnan

1.1 Project Rationale
Given the strong affinity towards travel among significant tech-savvy Gen Z members, they
are the majority users of digital travel services (Thi Loan & Viet Hoang, 2023). Hence, the
key users were further defined from the Gen Z demographic. Growing up in a digital world,
a vast majority primarily relies on smartphones for most of their travel-related tasks,
including pre-, during, and post-travel activities (Ozdemir-Guzel & Bas, 2021). Planning is an
essential stage of pre-travel, as the rest of the experience depends on it. It encompasses
the burden of expectation-related emotions and planning an itinerary that can meet those
expectations (Nichols, 2022). Queries regarding destination, transportation,
accommodation, meals, activities, and local customs and aligning them with the set
expectations can be daunting (Fesenmaier & Jeng, 2000).

Additionally, group travel adds the challenge of resource sharing and coordinating, which is
often overlooked in studies that focus on the marketing and influential aspects of travel.
Considering that many people prefer to travel in small groups and rely on asynchronous ecollaboration through available applications for travel planning (Thi Loan & Viet Hoang,
2023), it is crucial to consider planning from a multiple-user perspective rather than a
8

single-user perspective (Zhang & Sun, 2016). Navigating an overwhelming volume of online
resources and coordinating with peers is a challenge (Zhang & Sun, 2016). This challenge is
not unnoticed, and emerging technologies have introduced applications designed for
collaborative planning like Troupe and Wanderlog, with some also integrating AI to
streamline the process. Though both applications let multiple users edit itinerary they do
not cater to group planning at discussion and decision-making levels. Hence, solutions
today have surface level features with ample room for improvement, making it a ripe area
for design innovation.

1.2 Research Questions
●

What are the characteristics, behaviour patterns and mental models of the key
users?

●

What are the challenges faced by these people while planning a group vacation
remotely?

●

How can we design to streamline the vacation planning process (finding relevant
options and decision making) for the remotely connected group?

●

What features should be part of an application that can assist in group travel
planning (finding relevant options and decision-making)?

1.3 Objectives
This study aims to conceptualise a collaborative trip planning experience enhanced by
artificial intelligence (AI) tailored to specific needs of the defined users. Furthermore, it
streamlines the process of decision-making while planning and travelling.

9

Figure 1 Venn diagram representing objectives.

1.4 Scope and Limitations
This research follows an exploratory design process using design research, prototyping,
and testing. It focuses on identifying patterns and relationships between group travel
planning behaviour and remote collaboration among the defined users to conceptualise a
solution. To limit the scope of this study the key users had been defined as techsavvy, Gen Z individuals who frequently or occasionally enjoy traveling within a
small group. Moreover, the insights are based on primary research conducted with a
subset of ECUAD (Emily Carr University of Art & Design) students meeting this profile. Due
to the limited participant pool, this study only captures a segment of users, thereby limiting

10

the generalisability of the solution and rending the findings provisional. It is important to
note that this limitation is intended only to delineate the scope, rather than to exclude.

To meet evolving users’ needs, this concept would ideally use artificial intelligence. While a
lot of us might see AI as a tragedy for humans, there is an undeniable pace of innovation
within AI that continues to grow. The study recognizes the threats like the bias (Harding,
2022; Heaven, 2020) and risk of unethical surveillance but also the vast potential for a
positive experience when implemented safely. This includes using AI to analyse large data
and solve complex problems quickly.

This project focuses on proposing a positive experience for travellers planning in a group
using AI, without delving into the technicalities. We are aware that AI technology is not
flawless, and we are still exploring its capabilities. This research does not involve the
creation of actual AI systems; rather, it offers foundational thoughts on how to build them
responsibly in the future. The goal is to ensure that when AI is integrated into the proposed
solution, its operation is transparent, users consent to its use, and it is safeguarded against
misuse and privacy violations. AI should contribute to societal benefit, not just profit.
Although the creation and testing of AI algorithms is not involved as a part of this study, it
could still be helpful to generate a machine learning framework. This project’s UX
specifications can act as guides for creating the algorithm for recommendations or
suggestions (Kore, 2020). The rapid evolution of digital technology may render some
aspects of this research outdated quickly.

The prototype centres around the key features of an AI-driven collaborative travel planner
like personalised AI recommendations from social media posts, facilitating decisionmaking, route optimisation & collaborative planning. As a result, the prototype only
provides a specific set of features unique to this project rather than a comprehensive
mobile app experience. The study does not advocate the use of social media in any manner
11

for the use of the proposed solution, it is rather tailored to cater to the existing user base
that enjoy travel inspirations from social media.

Given the growing need and potential for improvement in current digital tools this research
aims to leverage design expertise to propose a more personalised and inclusive group trip
planning experience for primary users using AI capabilities.

1.5 Structure & Methodology
To design a solution according to the principles of design (Framework for Innovation - Design
Council, n.d.) the initial step emphasises understanding the audience, which in this case is
the tech-savvy, Gen Z individuals who travel within a small group. As a next step secondary
research was conducted to explore the behaviour and travel trends among Gen Z
demographics. Subsequently, primary research was done to validate the assumptions and
get a deeper understanding of user needs and problems. Due to limited research on group
travel planning and remote decision-making among Gen Z users, the demographic context
is from recent articles and blogs.

Concurrently, technological capabilities that assist in travel planning were explored by
experimenting with ChatGPT. Additionally, a comparative analysis of the existing travel
planning solutions was conducted to get a better understanding of current offerings and
user pain points.

In a later phase, participatory research was conducted with the understanding gained from
existing behaviour to dive deeper into the nuances of each step of planning before, during,
and post-traveling. However, since these interactions only provide a glimpse of the group
planning dynamics, an auto-ethnographic observational study was also conducted to add
insights from an actual travel scenario.

12

Synthesis of research led to the creation of two personas, a journey map which served as
constant reference while designing. The persona and journey map helped to visualise the
insights into a storyboard scenario of trip planning.

An iterative design process was followed with continuous creation and testing of mock-ups
to refine the solution, adhering to the approach of iterative design for optimal results
(Nielsen, 1993).

2. Setting the Context (Secondary Research)
This section sheds light on demographic behaviour and preferences, emphasising the
reliance on social media and digital technology for most of the tasks, including trip
planning. Beyond analysing this broad demographic for behavioural clues, a profound
inclination towards a mental model that accommodates flexibility in trip planning is also
discussed. Additionally, it highlights how AI is becoming increasingly integrated into our
daily lives, despite its risks. Lastly, a comparative study of the group trip planning
applications available today provides an understanding of the current market landscape
and their limited features.

2.1 Demographic Behaviour & Preferences for Trip Planning
According to the Oxford dictionary Gen Z are “the group of people who were born between
the late 1990s and the early 2010s, who are regarded as being very familiar with the
internet”. Growing up in the digital era, a large majority rely on smartphones for most of
their day-to-day tasks (Mason et al., 2022). These could range from reading the news,
ordering groceries to finding discounts on their favourite brands. It is therefore typical for
many of them to rely on their existing pattern of searching thoroughly through digital
resources for the best deal when planning a trip as well (Thi Loan & Viet Hoang, 2023).
13

However, with the involvement of social media, the usage is not limited to browsing. In fact,
social media platforms like Instagram, Facebook, and Tik Tok have a multifaceted and
significant impact in the realm of trip planning (The Path to Purchase, 2023). One of the
downsides is the phenomenon of FOMO (Fear of Missing Out) which compels some people
to travel without it being a personal desire by looking at others on a vacation. On the
brighter side, social media platforms help many youngsters to access relatable and
authentic content about a destination that they may like to visit or already intend to visit.
Hence, based on the situation social media serves as a resource or an inspiration.
According to a study by Expedia Media, about 77% of people rely on social media for travel
inspiration. Instagram, YouTube, & TikTok are the most popular applications used for
inspiration (Gonzalez, 2024). An increasing number of people are preferring to explore offbeat locations, due to easy access to information from social media influencers. Given the
Instagram behaviour of people saving posts and videos that they aspire to travel or add
activities in their bucket list, wouldn’t it be nice to have an itinerary directly created from
those assorted inspiration folder? Replicating the feature or adding additional inspirational
post links would just be an additional step. Hence, this was an important consideration for
seamlessly connecting people inspirations to their planning phase after validating the
similarity of behaviour with the participants.

A survey by Student Beans revealed a group of Gen Zs’ average spending per purchase
decreased for fashion (7%), tech (6%), and food (12%), while their travel purchases surged
60% (Roeschke, 2023). This is also reflected in the below figures that show 52% of them are
frequent leisure travellers. While they have a variety of reasons to travel like attending a
concert or event, adventure, or meeting family, the significant majority travels to relax (The
Path to Purchase, 2023) Being tech-savvy, they are often considered to be the most aware
mass in terms of taking care of mental health and physical health. Hence, many travel
being aware of the travel benefits on their mental and physical health (The Gap between
Traveler Expectations and Industry Perceptions, 2023) but at the same time, they also
14

understand the impact on the environment. So, many of them often choose sustainable
and environmentally friendly options considering the well-being of the planet (“Gen Z in
Travel,” 2023).

Figure 2 Data visualisation of travel frequency & motivations.

15

2.2 Experiences Enhanced by Artificial Intelligence
Majority of us, use some or other online service to get from point A to point B in our day-today lives. Without a Google map which optimises routes, predicts traffic, and acts as a
personal navigator we would be clueless. We are all aware of Google’s accidental collection
of more than just location data with the street view cars. But after having realised the
mistake, they immediately took the substantial steps to protect people’s privacy and create
a positive experience. More and more organisations are considering risk factors before
employing AI in their products. Provided AI is developed within ethical boundaries, we can
all agree that such intelligent services simplify our daily lives. The term artificial intelligence,
or AI, is a very broad term describing the power of computer systems to carry out
operations that would typically require human intelligence.

Modern AI, viewed through the lens of task-specific algorithms, amalgamates diverse
algorithms to accomplish a set of goals as described by Kore. For instance, a travel
application might integrate a reinforcement algorithm to tailor recommendations drawn
from historical data, employ Natural Language Processing (NLP) to decipher human
communication, or utilize vision algorithms for image data interpretation of places. These
systems fall under the segment of narrow AI, characterized by their proficiency in
dedicated tasks and distinguished from the theoretical constructs of super intelligent AI,
which are said to surpass human intelligence across all facets—a theory that has stirred
considerable discourse and threat among researchers. This study follows the ‘modern AI’
notion of an intelligent or rational agent seeking the optimal result even in the event of
uncertainty. Further explorations and designs cater to this understanding of modern AI
(Kore, 2020).

In the realm of travel, such algorithms are already used to provide personalised
recommendations for flights, stays, and tours based on past travel history, preferences,
16

and behaviours (The Impact of AI on the Travel Industry | McKinsey, 2023). The real-time
updates on weather, traffic and flight delays have allowed us to plan on the go.

By now Artificial intelligence is so deeply ingrained in our lives (Harding, 2022) that we only
realise its importance when the internet goes down. It has made our interactions not only
for travelling but also for routine activities like working, banking, shopping, and socialising
efficient and easier. Progress in this field is seen to be growing making it inevitable and
obvious to consider for this study.

AI-based Chatbot assistants
Chatbot also known as conversational agents are digital assistants created to interact with
humans in real time. These can be categorized into two types: Rule-based chatbot and
Advanced chatbot (Mucci, 2024). A basic rule-based chatbot used for simple tasks like
customer service operate on defined parameter and pre-set responses. Whereas an
advanced chatbot driven by AI has a capacity to mimic human conversational patterns,
solve problems, respond dynamically, and improvise based on past interactions.

Within Slack channels, for instance, such AI-driven chatbots have been instrumental in
summarizing long conversations and alerting user to mentions and reminding for
scheduled meetings. Similarly, social networking chatbots on WhatsApp and Snapchat can
be added to a group for problem solving or coming to a consensus. Most of these chatbots
are only activated upon user’s request. However, emerging research has laid the
groundwork for having proactive chatbots that can facilitate group discussions by
considering diverse opinions, promoting equal participation, and facilitating a non-biased
decision (Kim et al., 2020). Remote group decision making for travel can highly benefit from
such an AI-based chatbot facilitator.

17

Presented below is a table that captures the comparison of abilities essential for precise
and effective facilitation for group trip planning between humans in the group, human
travel agent, rule-based chatbot and advanced chatbot that uses AI.

Scale
High - Indicates facilitator performs exceptionally well, with advanced capabilities for this
task.
Moderate - Indicates facilitator performs adequately in this area, with abilities that meet
basic requirement.
Low - Indicates facilitator performs the task with limited scope and effectiveness.
Variable - Indicates performance to vary based on individuals or situations.

Abilities

Personalised
Interaction
Understanding
of Complex
Needs
Real-time
response rate
24/7 Availability
Problem
Solving skills
Ability to
handle large
data
Personalised
suggestion
Multi-language
Support
Promote equal
& diverse
opinions

Human
Facilitator from
the Group
High

Human
FacilitatorTravel agent
High

Rule-based
Chatbot
Facilitator
Low

Advanced
Chatbot
Facilitator
High

High

High

Low

High

Moderate

Moderate

High

High

Low
High

Low
High

High
Low

High
High

Low

Moderate

High

High

Variable

Variable

Low

High

Variable

Variable

Moderate

High

Variable

Variable

Low

High

18

2.3 The Balance Between Pre-planning and Spontaneity
This study does not assume all vacation types demonstrate the same decision-making
pattern, rather it builds upon a conceptual framework that suggests the decision-making
depends upon the characteristics of trip and traveller (Hwang, 2011).

Figure 3 A conceptual approach to travellers planned and unplanned attraction visits (adapted from Hwang, 2011, p
289)

This study explores the planned and unplanned aspects of the trip drawing from the
enroute theory of Hwang and mental models by Indi Young. Hwang suggests that people
usually anticipate their planning with insufficient or incorrect information available to them
and then revise it based on the enroute conditions that arise and new information
collected while travelling. While these decisions are highly dependent on the traveller’s
characteristics, measuring the unplanned travel decisions is still under research (Park &
Roehl, 2016)

Travel planning from a mental model perspective became necessary, to better understand
the thought processes and mindsets of people during planning. We generally have a
sequence of tasks in mind for approaching a problem based on our past experiences which
can be visualised through a mental model (Young, 2008). In the case of travel planning
19

many factors come into play like the personality of the person, level of comfort with
spontaneity, tolerance with risk and logistical preferences. Moreover, a person who is
spontaneous majority of the times, depending on the situation may preferred planned
things and vice-versa.

Acknowledging the gaps in the literature regarding mental model for travel planning, this
study seeks to identify the patterns of travel planning process through participatory
research.

2.4 Remote Collaborative Planning
Most of the communication for Gen Z with friends primarily occurs through texting on their
phones. They are also efficient at collaborating remotely for work and education,
demonstrating a comfort level with digital tools and remote means of communication.
Given their preference for texting and their experience with remote collaboration tools, it is
reasonable to deduce that they would also likely use a digital tool to plan their trips
remotely. The remote collaboration planning for a trip would include several tasks,
including the planning of the schedule, transport, accommodation, the choice of activities,
and others. This would also entail considering the diverse opinions and coming to a
consensus. This definition of collaborative planning is similar to Healey’s description in
“Collaborative Planning: Shaping Places in Fragmented Societies” (1997) which also
highlights shared ownership and inclusion in planning. However, planning collaboratively
while being remotely connected is still a notable challenge for many (Gonzalez, 2024)

In the case of remote work, efficiency has been improved by asynchronous planning and
automating some of the tasks like scheduling meetings, setting reminders, and assisting in
drafting documents. In recent times, progress has been made with AI analysing large

20

volumes of information to provide insights for informed decision-making, crucial in remote
teams.

Despite considerable young travellers holidaying with their friends, very few studies have
attempted to explore the topic of shared decision-making among groups of friends
travelling (Decrop, 2005). Therefore, the understanding of digital collaborative planning is
derived from the models of remote work environments and is further investigated by
means of an examination of the collaborative trip planning applications that are now in
use.

2.5 Comparative Analysis
Comparative analysis was carried out to understand the current user experience offered by
applications serving the need for collaborative travel planning. Primarily an exhaustive list
of all trip planning applications was created and classified into two groups: direct
competitors or those focused on group travel planning, and indirect competitors, or
applications related to planning in some way. From a pool of 11 direct competitors, three
were further shortlisted based on popularity and unique features. Among the three were
Troupe, MiTravel, and Wanderlog. The aim was to study the features, and their relevancy
and derive valuable insights from the user reviews. A thorough evaluation was conducted
on each of the following factors: services offered, pricing, value propositions, establishment
dates, reviews and ratings on the App Store & Play Store, features, the tone of the copy,
and the overall user experience.

21

Discovery and Analysis

Figure 5 Group collaboration
with voting.

Figure 4 Smart search with AI
Assistant

Troupe
Troupe is a collaborative planning application launched in 2020 that places a strong
emphasis on group coordination. Its standout features encompass the ability to make
suggestions, conduct polls, facilitate voting, and make decisions. The desktop version
further integrates an AI assistant named Troupie, enhancing the user experience with map
views and accommodation recommendations. However, this functionality is currently not
available in the mobile application. While the application is user-friendly, its usability is
hindered by a relatively high number of required clicks for navigation and planning. The
manual process of adding places to the itinerary contributes to this challenge.

22

Figure 8 Collaborative Itinerary

Figure 7 Search with AI

Figure 6 Resources from social
networks and blogs.

Wanderlog
Wander log, launched in 2019, offers a comprehensive solution for vacation planning with a
strong emphasis on user-friendly design. This all-in-one app allows users to plan & track
their travel plans, explore new destinations and accommodations, and exchange travel tips
with friends and fellow explorers. Wanderlog covers everything from collaborative itinerary
creation to expense management. However, it lacks a clear focus on collaborative planning
tasks, such as discussing and negotiating options. Nevertheless, it excels in providing an
excellent map view alongside itinerary planning, making the planning process more
convenient. Additionally, the AI assistant, while a valuable feature, is not yet seamlessly
integrated to offer a fully personalised journey experience, requiring users to interact with
it separately.

23

Figure 11 Mood board style
itinerary.

Figure 10 Poll feature for
group decision making.

Figure 9 AI Assistant which in not
context aware.

MiTravel
Mitravel is a collaborative itinerary planning tool that is still in development and is squarely
focused on refining the collaborative planning experience. This start-up is currently in its
beta testing stage. Mitravel incorporates features like polling, voting, and note-taking, akin
to what Troupe offers. What sets it apart is that all these collaborative activities occur on a
virtual board accompanied by an integrated map, enhancing the planning process. In
addition, it offers a convenient 'search and add destination' feature, like Wanderlog,
simplifying the process of adding locations to the itinerary. It also has an AI assistant which
is again context aware and acts as a separate search engine.

Findings
Observing the continuous evolution of these applications in response to shifting market
demands and technological trends reaffirmed the validity of the problem space. The
consistent user reviews, all expressing a clear desire for an enhanced collaborative travel
24

planning experience, provided additional confirmation of the timeliness of such a solution.
Furthermore, the ongoing initiatives of established applications like Wanderlog and Troupe
to integrate AI features highlight the increasing significance of AI assistants in group travel
planning. However, it is important to note that these AI integrations in existing solutions
are not yet fully polished or seamlessly implemented; they often function as additional
features rather than integral components of the user experience. It is within this context
that the potential to design a group travel planning application that takes full advantage of
AI-technological advancements to enhance the user experience is recognised.

25

3: Primary Research
As a part of primary research a participatory research activity, Experiments with ChatGPT
the basis of informal interviews, and an auto ethnographic observation study were
conducted. For each method, the below text delves into the rationale behind its selection
and the specifics of how the research was carried out. This is followed by an analysis of the
findings, and a reflection on the methodologies employed, considering how they might
have been optimised.

3.1 Informal Interviews
Understanding people's expectations and problems with human travel agents was
essential in the initial stages of designing an AI-based travel assistant. It was also important
to determine whether AI would be able to address these needs. Hence, informal interviews
were conducted with 5 participants to understand their experience with travel agents and
identify opportunities for AI travel assistance and pain points.

Analysis & Discovery
Participants expected travel agents to plan and book travel arrangements as per their
needs and preferences. They preferred having recommendations for activities, tips, and
advice. Apart from that in case of any unexpected events like flight delays or severe
weather, they expected agents to navigate and reiterate the plan.

During the interviews participants showed openness towards having an AI driven travel
agent and the willingness to explore the potential.

Some of the common pain points highlighted from the interviews were as follows:
“Telling the Advisor about all your preferences.”
26

“Uncomfortable sharing personal preferences.”
“Incomplete advice on safe places to travel.”
“They don’t know you well enough to suggest the best option.”

Findings and Reflection
At this stage, the decision of moving forward with the AI as a solution for travel planning
was still in the air. Hence, informal interviews proved to be quite helpful for this study in
gathering user thoughts on travel agents rather quickly.

3.2 Experimenting with ChatGPT
With this understanding of what users expect out of a travel agent, experiments with
ChatGPT (AI based chatbot that uses natural language processing to create humanlike
conversations) were conducted. ChatGPT was assumed to be a digital travel agent and the
researcher acted as a potential user planning a trip.

Approach 1: A conversational approach where users can ask questions, put forward their
preferences and gradually refine their choice of a destination based on the suggestion
made by AI-chatbot.

Approach 2: Users can enter their preferences by answering pre-set questions which
would enable the algorithm to curate a perfect itinerary. Further modifications can be
made through conversation. (Note: In the example below the preferences are presented in
the form of a chat which would be collected beforehand.)

27

Figure 13 List of activities.

Figure 12 Budget Calculations

Figure 14 Detailed itinerary.
Note: Figure 12,13,14 and 15 are screenshots of conversation generated by the researcher using GPT-3,
OpenAI’s large-scale language-generation model

Approach 1:
The chatbot was prompted to suggest adventure activities in Okinawa, Japan. As expected,
it generated a list again with a description which helped to visualise that activity. (Refer to
Figure 13: List of activities)
28

Subsequently, the Chatbot was asked to formulate a comprehensive itinerary for a six-day
vacation to Okinawa. Despite initial reservations about its capacity to handle such a
complex task, the results were unexpectedly impressive. The AI-chatbot generated a
detailed itinerary, encompassing a variety of activities, suggested tourist spots, and culinary
delights for each day of the journey. The intricacy and accuracy exhibited by the AI were
noteworthy (Refer to Figure 14: Detailed itinerary). Furthermore, the AI was tasked with
calculating an overall budget for the Okinawa trip. It produced an efficient and
comprehensive estimate of the expenses as well. (Figure 12: Budget)

Figure 15 Itinerary based on preferences.

Approach 2
By clearly stating all preferences in a single request to ChatGPT (for testing purposes), it
generated a comprehensive itinerary, breaking it down into daily activities that include
29

recommended places to visit, food to try, and activities to participate in. It also ensured that
all preferences were considered be it of multiple people. (Refer to Figure 15: Itinerary
based on preferences). This method of obtaining the necessary information first and then
generating the itinerary based on it seemed efficient as compared to the previous one.

Findings and Reflection
It was evident from the experiment that conveying the needs to the system was the only
way to get the most precise outcome. The results accuracy depends highly on the input
precision. However, not all users may know exactly what to ask for, risking the possibility of
missing an important preference. Hence, approach 1 is taxing, both in terms of time and
cognitive effort. Having multiple outputs makes it difficult to track for the correct one. On
the other hand, approach 2 simplifies the process by offering pre-set questions for
preference selection, leading to a generated itinerary. As netizens, people have grown
accustomed to relying on prompts and recommendations for searching, so for a digital
travel agent to be effective, it should be able to guide the conversations and possess
contextual awareness.

3.3 Participatory Research
To understand the nuances of collaborative planning across groups of different numbers of
people, a participatory research study was conducted with 1, 2 and 3 people in a group.
This research activity also sought to identify the planned and spontaneous areas of the trip,
as well as its relevance.

The participants were given post-it notes with tasks written on it like accommodation,
travel, budget, activities, taking and uploading pictures, and other related items to be
arranged under one of the four travel phases: pre-planning, planning, while travelling, and
post-traveling. Questions related to each of these tasks were asked while the participants
performed the activity.
30

Preplanning is the stage before planning for the final spot, when the possibility of visiting a
potential spot is being considered.

Planning begins as soon as an agreement to travel is made and a particular trip is
discussed.

While travelling is when one has started the journey towards the trip destination until one
return home.

Post travelling is the period following one's return from the trip.

Discovery and Analysis

Figure 17 Result of activity with 1 participant.

Figure 16 Result of activity with 2 participants.

Figure 18 Result of activity with 3 participants.

31

The participants were unable to provide clear responses when asked to imagine a future
trip as they did not have any firm plans. Rather, they found it easier to respond to the
inquiries based on past experiences.

Despite the different numbers of participants, no notable differences were observed
between the planned and spontaneous portions of the journey. The top priority during preplanning was reaching an agreement on dates, followed by comparing flight costs to
determine a destination. While some preferred to seek inspiration as a part of pre-planning
even before creating a budget, others preferred to look for inspiration later somewhere
between pre-planning and planning.

Booking accommodations was a crucial component of planning that was desired by all
participants. Participants mentioned thoroughly checking prices with and without taxes,
comparing several options and negotiating for comfort. One participant described her
unsatisfactory stay at a budget hostel, mentioning the smelly shared restrooms and
uncomfortable bedding as reasons why her trip was less enjoyable overall. Based on the
Peak-end rule people remember the experiences at the intense point and end points
rather than each moment (Bassett, 2023). Having understood that People sought
accommodation that was more than a stay based on their previous experiences. Clearly,
accommodation played a role in creating memorable peaks by offering comfort and
amenities like a pool, private washrooms, Wi-Fi, and ease of access. However, while
planning, other aspects like internal transport, places to visit, and activities were
considered with a degree of flexibility, mostly left open to last-minute adjustments.

The prioritisation of planning accommodation over places to visit mirrors the principles of
prospect theory (Kahneman & Tversky, 1979) which states people are more driven to
prevent losses than to make gains. When planning a vacation, finding a good place to stay
is top priority because it is thought to be a means to avoid suffering a big "loss." A bad
32

hotel selection can significantly reduce the enjoyment of the vacation. So, making sure to
have good lodging is like not losing money. Planning events and attractions, on the other
hand, which are seen as "gains," is frequently more adaptable. This flexibility reduces risk
and allows for travellers to make impromptu decisions and changes based on their
experiences and current information while on the road.

While grappling with the overwhelming amount of information while planning, they
preferred AI to provide a shortlisted set of options rather than making the final decision.
Here are a few statements from the interviews that capture the emotion of searching
through a plethora of sources to discover the ideal option.

“It's a little overwhelming because it's like a lot of considerations. I would say that at least
because I mean I'm not in a position in my life now where I can pick any Airbnb “

“So, I'll pick like five and then I'll go and revisit them and be like, well, how much are they
actually? And so it's a second layer of filtering.”

“I feel like it's really helpful. I mean, it's overwhelming in some way. And then you have the goal
of finding out the details about this city so there's gonna be lots of reasons like that you get
through reading this post. Personally, I feel like less anxious about like reading it then just go
somewhere and have like no expectation”.

In terms of collaborating in a group, participants expressed an interest in negotiating and
compromising while still being able to do things that were favourable for them. In contrast
to the individual candidate, who was more focused on their personal checklist, those in
groups of two and three were more considerate of others' opinions.

33

Findings
The primary finding was a pattern in the dynamic model of planning which included
collaborating and booking for high-stake elements like accommodation and transport but
allowed for flexibility for places to visit and activities. Additional results from this exercise
supported the demographic data that using multiple applications to plan a vacation is
common (Expedia Media). Lastly, a participant travelling the next day mentioned about her
stress of committing to the school activities along with preparing for the trip. This
highlighted the challenge of balancing daily chores with the demands of planning a trip.

Reflection
In a mock round conducted with a participant, questions were asked chronologically from
the prepared list. It soon became apparent that this approach was interfering with the
participant's flow of thought while putting the post-it notes. So, the format was improvised
to offer questions based on the participant responses. This way they were able to provide
specifics and details to the scenario. However, this made it challenging to remember which
questions were answered and which ones were not. Nevertheless, this approach was
chosen as the finer details provided a better understanding and gave depth to the
response.

Responding became challenging for participants when asked to envision a scenario that
entailed the consideration of numerous factors, thereby imposing additional cognitive
strain. Which is why the participants found comfort in answering based on a lived
experience. Thinking of a previous trip, contradicted the idea of travelling with the other
participants. Hence, this activity did not yield many insights into group dynamics while
planning.

Instead of having multiple areas of intervention this activity could have been split into two
parts. The first part to understand the planning journey encompassing the planned and
34

spontaneous areas. While the former part could have been aimed at group dynamics using
a specific situation and tasks.

3.4 Auto-ethnographic Observations
During a trip involving four travellers, including the researcher, auto-ethnographic
observations were conducted from the stage of pre planning until post travelling. These
observations were documented on Google notes and further expanded after the trip. To
ensure integrity of the research no conversations or events were altered, and the travellers
were not informed of the research activity. The observations were made taking the
necessary precautions. No conversations or situations were manipulated for the sake of
research. This approach aimed to capture authentic experiences and insights into the
travel process. To do so complete anonymity of the travellers' identities is preserved
throughout the study.

Discovery and Analysis
While being in distinct parts of the world with different time zones, finding mutual time to
communicate was a challenge. The entire planning until the day of travel carried on for 2
months during which very few calls were made with everyone's presence. The longest calls
were for discussing the initial idea of the trip and booking accommodation. Otherwise,
most conversations took place asynchronously over text.

Booking.com, Expedia and other platforms were used to search accommodations and
flights which were then shared on the group for discussion. For accommodation there were
different preferences that came up in terms of budget, facilities, ambience, and location.
After a lengthy process of negotiation and multiple discussions, two accommodations were
selected. One with a lower budget and another at a prime location with facilities were
booked for different halves of the trip. After booking the flights and accommodation more
than a month in advance, planning the itinerary was brought up multiple times but no
35

initiative was taken to do it with the group. Individually, people would take inspiration from
YouTube videos, Instagram posts and share links on the group.

While travelling in a group, individuals openly expressed their preferences and were
accepting towards taking up different activities based on their preferences. On a particular
day while some travellers enrolled in an adventure activity the others enjoyed a restful day
at separate locations. All of them got together at the end of day to grab dinner and
expressed enthusiasm to learn about each other's experience.

However, in another instance when the group unequally split on a plan to socialise with
new local acquaintances, the individual with difference in opinion seemed to have been
under pressure to move forward against the will since the majority aligned on the plan.
Although initially resistant, acknowledging the group's collective preference led the
individual to eventually agree with the plan. This instance reflects the complexities of group
dynamics, especially when decisions do not align unanimously within the group.

Findings
Multiple applications were used to plan and make booking aligning with the former
collected the data and participatory research. When planning a trip, certain choices such as
accommodation and transportation require consensus among the group. Conversations
become difficult when individual preferences are different from those of the group. Such
conflicting situations seek a non-biased approach in decision making. The overall planning
journey consisted of the planned and flexible in alignment with the participatory research
findings.

As a result of the primary research, it can be established that people are exploring easy
options to plan. Besides the literature (The Path to Purchase, 2023) interviewed participants
and autoethnographic observations also reinforced the use of multiple digital tools for
36

planning in advance and making changes as needed. Similarities were spotted in terms of
the planned and flexible mental models of the trip. Behaviour patterns such as
asynchronous planning and the challenges while communicating remotely were
highlighted. Nevertheless, an established understanding of individuality and diversity of
preferences led to openness to different plans.

4. Synthesis
4.1 Insights
Overall, the digital dependency leads to the need for more convenience and efficiency in
the planning processes, participants frequent use of numerous applications and platforms
was a clear indication of their digital dependence. The reliance on technology has made
complex tasks like keeping track of the bookings and sharing resources across the group
much easier. However, unmet needs like hyper personalised search results and effective
asynchronous collaboration drive the demand for innovation. The primary research
provides valuable insights into such areas that are crucial for remote group trip planning.

Demand for Hyper Personalised Content
The main source of inspiration being social media, reflects participant’s preference for
personalised content and authentic information. They follow travel influencers that align
with their values and preference, the algorithm analyses this behaviour and suggests more
posts like that. This process enables participants effortlessly discover relevant and
authentic content. The comparative analysis of trip application also points towards the
increasing usage of AI to get personalised itinerary and travel recommendations.

Diverse Preferences - Respect for Individuality:
37

Participant’s tolerance for a wide range of preferences reflects their regard for personal
uniqueness. They were open to the notion that members of the group may choose to
engage in different activities while on a trip and values diverse experiences. Their
acceptance of diversity and appreciation of differing interests makes for a more adaptable
vacation experience, acknowledging that being together all the time is not required.
Though they acknowledge diverse preferences they are firm about accomplishing personal
interest and preferences.

Asynchronous Collaboration - Adaptability to Varied Schedules:
Planning a trip for the participants was not an isolated task. It existed amidst numerous
daily responsibilities ranging from academics to household chores, part-time jobs to social
interactions, and keeping up with trends. Many also plan trips with friends located in
various parts of the world in a different time zone. Their multifaceted lifestyle and cross
time zone connectivity emphasise the necessity for asynchronous communication methods
that work well with their hectic schedules. As a result, travel planning can be a more
inclusive and stress-free experience since they can participate at their convenience.

Balanced & Effective Decision-Making Process
Conflicting scenarios where individual preferences are different from those of the group,
often pose a challenge as per observations. Such situations seek for a non-biased approach
in group travel planning wherein multiple perspectives are considered equally. Remote
planning also calls for considering delayed, asynchronous cognitive processes to support
some planning decisions. This would also be an advantage for those, who find it
challenging to participate in a real-time debate.

Dynamic mental models
A dynamic planning approach with a flexible mental model, willing to accommodate any
modifications was favoured by all participants. Even while the itinerary may have certain
38

predetermined elements, there was still sufficient flexibility to allow for unforeseen events
or modifications. This mental model was common among the results of participatory
research and auto ethnographic observations, which indicate that the pre- and duringtravel stages consist of a combination of structured and flexible components that
accommodate the group's varied and changing preferences. Based on the analysis of
primary research the tasks involved pre travel and during travel were categorised under
“Definitely Planned” and “Flexible” mental models of the participants.

Figure 19 Definitely planned and flexible areas of the trip represented with post it.

4.2 Persona
Based on the results of primary research, personas depicting the two archetypes—the
Instagram-Savvy Scenic Lover and the Budget-Conscious Explorer —were created. The
purpose of this action was to feed the creation of a scenario that may function as a

39

convenient point of reference for the design of application.

Persona 1: Amanda, The Instagram-Savvy Scenic Lover
Figure 20
Persona 1:
Amanda,
The
InstagramSavvy
Scenic
Lover

40

Persona 2: Rachel, The Budget-Conscious Explorer

Figure 21Persona 2: Rachel, The Budget-Conscious Explorer

41

4.3 User Journey Map & Scenario
Taking into account the nuances of group trip planning involving different preferences, a
journey map was created. This map visualises a scenario in which Amanda and Rachel two
friends who collaboratively plan a trip together. The journey begins from the inspiration
phase until they start travelling. Each phase expands further into the relevant touch points
based on the activities involved and emotional ride throughout the communication and
decision making. Potential design opportunities for each phase were identified through this
comprehensive process.

42

Figure 22 Journey Map of Amanda and Rachel.

43

Scenario of navigating conflicting thoughts regarding accommodation
Imagine this scenario, where two friends Rachel residing in Toronto and Amanda in
Vancouver are planning a five-day trip to Bali. They have already gone through the phases
of inspiration and serious consideration during which they booked the flights. They have 25
more days until the trip, and they are now considering stay options. In the below scenario
Amanda and Rachel hit a conflicting point, specifically concerning budget, amenities, and
views while discussing accommodation. Amanda prefers to economise on accommodation,
opting for a less expensive stay. In contrast, Rachel prioritises comfortable stay with
facilities and a scenic view. A context aware AI chatbot called Zeelo who is also a part of this
conversation analyses their needs and helps them navigate the situation with ease.

The algorithm, being trained on best approaches to navigate group decision-making,
suggests ranking from the identified factors concerning users for choosing an
accommodation option. This way users were able to express preferences in a more
detailed manner, not just their top choice and avoid polarised outcomes. The respective
features of the chatbot are highlighted by the number and explained later.

44

Figure 23 A Storyboard of AI chatbot facilitation through an example.

1. Relevant and personalized recommendations
• Offer recommendations based on the group's expressed interests and past preferences.

2. Analyse Group Needs, Problems & Conversation Dynamics
• Collect & analyses data on keywords and phrases related to travel planning, like
"accommodation," "time," "location," etc.
• Tracks the frequency and sentiment of each participants contributions.
• Determines the main topics of discussion and conflicting opinions.
• Leverages analysis of previous chats to understand group interests and common
decisions.

3. Promote Equal Participation
• Detects less active or influential group members based on their participation level.
45

• Direct questions or prompts to less active members to encourage their input to foster
inclusive conversation.

4. Facilitates decision-making
• Suggest an optimal path forward based on learnings from research papers regarding
approaches for effective decision-making in small group.
• Summarizes the points made so far to ensure everyone is on the same page.

5. Maintains Transparency
• Is transparent about the way it operates
• Gives credit and ownership to relevant sources it was trained on or takes inspiration
from.

6. Privacy Focus
• Ensures confidentiality of user responses.

5. Outcome
5.1 Ideation
The ideation phase involved unrestricted brainstorming, a common approach in design to
foster creativity. However, this led to digression from the core focus, resulting in looping
into designing the onboarding steps. A critical turning point occurred while presenting it to
a participant, who questioned the role of AI in facilitating collaboration. This feedback
highlighted the significant omission in the design outcome, compelling for a refined
approach. By focusing on a clear problem statement “How can AI assist in collaborative
planning?” the process became anchored ultimately leading into the chat format-design.
46

The initial attempts consisted of designing the flow of onboarding and adding saved social
media posts for a personalised itinerary to begin with and choosing the location and ideas
around how that can be streamlined. The main intention with using social media saved
posts as a channel for inspiration was to enable end users to influence the ML-generated
suggestions based on their own interests. Some of the shortcomings that have been noted
for AI products include bias and irrelevancy in data and unreliable results. To address this
issue, the idea was to enable users to create their pool of data from social media.

For the next attempt, a journey map was created with 2 different personas for the scenario
of them travelling together. Instead of the entire journey the focus was on choosing an
accommodation. This was also what most of the participants emphasised on being
particular about and was a matter of discussion in the auto-ethnographic observation in
terms of the budget and facilities and view. The ideation phase involved thinking about
ways in which an AI assistant would help to navigate difference of choices in an unbiased
manner. This led to the chat-based approach as seen in the image.

47

Figure 24 Ideation sketch of a chat-based approach.

48

5.2 Wireframes

Figure 25 Adding options for places initial attempt.

The first screen outlines the design for list of trips as a user can plan more than one trip.
Upon selecting a trip, the user would land on a recommended itinerary. (note: it was
considered that the user already undergone the process of creating recommended
itinerary). The goal was to design a user-friendly interface for enabling users to add more
than one option through exploring. In the initial design, the search feature expanded on
the same page below the destination as seen on the last two screens. However, that that
seemed to clutter the screen unnecessarily with a higher risk diverting the users focus from

49

the primary task.

Figure 26 Adding options for places revised for a better experience.

Hence, this flow was reimagined with a drawer. Wherein upon clicking “Add Options” the
users would be able to search and explore through a range of options without any
distractions in a drawer and easily add these options. The drawer also assures the users,
that they can easily access the itinerary just by tapping out. Upon adding the options, a
reminder to set time a limit to vote between options pops up (screen 3). This would help
users to effectively decide in each timeframe. The last screen is a representation of
multiple options and votes once users have undergone the adding process.

50

Figure 27 Feature considerations for a chat-based solution: AI-facilitation for choosing accommodation.

These are the wireframes for a chat-based flow of AI- facilitating to decide accommodation.
In this scenario the AI-suggests users to rank the shortlisted options to reach a consensus.
The features and interactions that were thought of include allowing users to rank privately
which is seen in screen 3 and 4. Another feature consideration was for muting AI seen in
screen 5 when users do not want the AI to interfere in some conversations.

5.3 Final Concept
The final concept is a set of features and user flows that focus on the insights obtained
from this study. These cater to the following three areas:
1. Personalising the itinerary recommendations for the group using AI, based on
shared social media posts by individuals.
2. Ease of modifying itinerary by adding, removing, or reordering travel destinations.
3. AI-facilitating decision making for navigating complex conversations.

51

Relevant Recommendation

Figure 28
prototype
design
representing
flow for
relevant
itinerary
generation.

Link to view prototype
As per the insight’s the users wanted hyper personalised recommendations and
consideration of diverse perspectives of group members. With AI, personalisation of the
output depends on the accuracy and relevancy of data. Since, the findings from the
participatory research revealed that for the participants, social media already acts as their
primary source of inspiration; therefore, leveraging data from content creators of social
media to create a personalised itinerary would streamline the planning process. By
integrating the social media access feature, the aim was to align with established user
behaviour rather than introducing additional, unfamiliar steps. As depicted in the figure,
52

once users grant permission, they can select specific saved posts of other content creators
that they find inspiring. These posts would be saved in a personal folder that can be
modified by the user as needed. Once all the users upload their inspiration boards, the
application then analyses these posts using NLP (Natural Language Processing) techniques
and vision algorithms. The AI system learns what types of destinations, activities, or
experiences are commonly preferred among the group and suggests similar options. The
core ML component would be a hybrid recommendation engine. It would use multiple
methods like content-based filtering, collaborative filtering, knowledge-based, and social
media-based recommendations, to process this data from a vast dataset (source not
defined but could be authorised and user consented google reviews, booking apps like
Expedia, Booking.com, TripAdvisor, destination site, etc). A hybrid recommendation system
with a weighted influence from the added social media posts of users over other data
would be more effective as it addresses the limitation of individual methods. This would
eventually create a personalised itinerary for the group.

53

Collaborative Planning

Figure 29 Features for ease of collaborative planning.

Link to view prototype
The itinerary generated with multiple recommendations can act as a starting point for the
group to further collaborate and iterate according to preferences. This collaborative
planning includes the ability to add, choose, and vote on shortlisted options. The search
encompasses AI-powered personalised recommendations along with the typical manual
search. Users can easily add or remove locations and seamlessly rearrange them within the
itinerary. The interactions are designed keeping in mind the current behavioural trends to
minimise the learning curve. Additionally, a map view is available to facilitate the planning
process, allowing users to visualise and coordinate their chosen destinations with ease.

54

AI-Facilitator

Figure 30 Screens showing AI-chatbot facilitating group discussion for deciding accommodation.

Link to view prototype
The comparative analysis findings show that while people are willing to plan, they are
seeking digital tools that simplify the group planning process altogether. As texting was
identified as the most favoured means of communication from primary research, it would
make sense to choose an AI-powered chat feature to significantly enhance the planning
process. Not only can it provide recommendations for places based on the past
conversations but can also facilitate decision-making, promote diversity of opinions and
equal participation to resolve conflicts. As per the insights, making an unbiased decision
would involve consideration of diverse perspectives. With the capability of sentiment
analysis AI would be able to consider diverse perspectives from an emotional lens,
recognise the tone of conversation, and promote equal participation (Kim et al., 2020).

The above screens are designs for the application are representative of the Scenario of
navigating conflicting thoughts regarding accommodation discussed in section 4.3. A
55

context aware AI chatbot called Zeelo, helps to decide for the most favourable
accommodation. The participants encountered a disagreement about accommodation
choices, specifically concerning budget, amenities, and views. One person in the group
preferred to economise on accommodation, opting for a less expensive stay. In contrast,
another individual prioritised comfortable stay with facilities and a scenic view. The
algorithm being trained on materials with best approach for to navigate group decisionmaking for trip would suggests ranking from the identified factors concerning users for
choosing an accommodation option in this example. In a real case, it would depend on
multiple factors like historic conversations of the users, shared inspirations, conversation
dynamic based on which AI will analyse and suggest a suitable approach.

5.4 The Working Explained
This diagram, typically called a System Architecture Diagram, describes how the
technological aspects of the project work together to create the user experience.

56

Figure 31 System architecture diagram.

57

5.5 Testing and Iterating

Figure 32 iteration screens of collaborative panning and AI collaboration.

58

Testing was carried out multiple times as part of this study, and changes were made to the
features and user flow in response to feedback. Users were given an overview of the
solution and were instructed to think aloud while testing. This served as a research method
as well and yielded valuable insights on the user flow. In the initial 2 rounds the users were
having a tough time figuring out the flow and the concept seemed unclear. The main query
revolved around the role of AI in facilitating. Thus, the third iteration explored a chat-based
approach using a relatable scenario for selecting accommodation, which was highly
appreciated by the users. The feedback on this concerned whether it was possible to stop
suggestions while travelling? This led to the consideration of an important feature to mute
or pause the AI assistant whenever the interference is not pleased.

Six people volunteered to interact with the final prototype at the open studio event where
it was displayed. The overall response with using AI for group trip planning was positive.
Majority of them were intrigued by the solution and showed curiosity towards a full
experience. While the initial idea was validated before, the final testing allowed for the
validation of the comprehensive concept consisting of the user interface and features.
Some individuals suggested extending the app's capabilities to incorporate a more
extensive range of digital functions for travel in groups, like photo sharing, cost splitting,
and booking.

However, adding these may drift away from the core focus of the app, which is to assist in
decision-making among a group. When an application tries to do too much and shifts from
its primary intended use, it becomes less effective. Hence, no changes were made in this
regard.

59

6. Reflections and Future Directions
6.1 Reflection on the process
Overall, the iterative design process of constant testing and refining proved to be effective,
with rapid identification and design revisions. However, it is important to note that the
autoethnographic findings are likely to be influenced by researcher's unconscious biases,
limiting their generalisability beyond this study.

Narrowing the focus by defining the users and challenge of group travel planning, the
solution became tailored, and complexities of AI-driven systems were unpacked. Though
the idea and the user flow were validated with designs, a thorough testing of AI-based
design concepts can only be achieved with a functional model. The rationale is that for AI
models to become accurate, considerable training and testing are necessary. Therefore,
the concept's potential could not be completely evaluated unless that was adhered to. Also,
since the participants for this study were limited to ECUAD students, the solution cannot be
assumed to work for everyone unless tested with a wider audience. In the proposed
solution the frequency of the AI assistant interrupting the chat, identifying a conflict vs
discussion, accurately streamlining a suggestion are all aspects that can only be tested with
a functional model. Hence, a key insight from this study points towards the necessity of
concurrent design and the development process to efficiently test chat-based AI assistant
products.

6.2 Designers and AI
This study also identifies a need for designers to have a deeper understanding of how AI
works. This is essential because AI directly impacts the experience users have with a
product. Interaction of humans and AI can be studied through observations or interviews.
60

However, it is challenging to curate an experience being unaware of the actual capabilities
and limitations of AI.

A general trend observed in the emerging Generative AI based applications for content
creation, travel planning, schedule planning is the reliance on prompt-based approach. This
can be an opportunity where designers can contribute by enhancing the user-friendliness
and extend the possibilities of these applications to meet the needs of users.
Understanding AI will empower designers to refine user flows and identify key tasks that
can be seamlessly integrated into features instead of mere prompts. Just like in Adobe
Photoshop a frequent task of removing or extending background is an AI feature along
with the ability to edit based on prompts.

The most popular tool for UX designer which is Figma is also seen to have AI based plugins
for many tasks like ideation, mind mapping and converting designs to code. The progress
and constant upgrades with AI, makes it apparent that soon design prototyping tools that
would emulate AI functionality would be required for UX professionals to test AI-driven
experiences.

6.3 Future Directions
Another revelation that emerged clearly during the study was the demand for AI based
solutions. People wanted ease of access to information and the free AI license has led to
the widespread adoption across various domains including using ChatGPT and Bard for
travel planning. An influencer promoting Bard demonstrated the usefulness and simplicity
of these technologies by asking for customised trip recommendations for Ahmedabad, a
city in Gujarat. The public's rising predisposition towards AI assistance was further
reinforced by the positive response in the influencer's post's comments section.

61

It cannot be denied that humans and AI working together can be unbeatable. For instance,
AI can process large amount of financial data and provide options for reliable investments.
Yet, humans with their intuition, expertise and market understanding can identify right
options. Similarly, for travel AI can help with hyper personalised recommendation and
decision making given the accurate data. In case of travel planning AI-based products users
demand relevancy and accuracy of recommendation, which can be achieved only with
precise data given to algorithm. So, in the case of Zway as well the user experience as much
as it is dependent on AI is also dependent on the users themselves and the way they
interact. The design experience can be refined based on feedback, however moulding the
generative AI experience will be driven by users.

The reason why AI can provide accurate output is because it has access to abundant
resources and its infrastructure allows to process through multiple layers. However, this
capability comes at a significant environmental cost, as AI operations, both during usage
and training, consume substantial amounts of power. In addition, the open license has led
to an increased experimentation and usage. Hence, looking at sustainable ways of training
and using AI as well as responsible usage of AI would be crucial to lower the environmental
impact and ensure its long-term viability.

6.4 Ethical Considerations
Having known that data is the main force driving a product like Zway, which is a subject of
concern for all users, building the algorithm with an ethical approach in compliance with
lawfulness and considering all the anticipated risks is a must. While providing a practical
solution it is imperative to ensure the users’ privacy is uncompromised. In the context of
users saving posts from content creators to use in a dataset it is essential to take consent
from the creators as they are owners of the data. This can be achieved by informing the
content creators about the use of their posts for AI data sets and giving them the option of
sharing anonymised data excluding any personal information. Enhanced techniques of
62

data masking can be employed to anonymise user data while preserving its utility for AI
training. This approach could serve as opportunity for content creators to monetize
without compromising their privacy. They could set a price for their own data for allowing
other people to use their data on Zway or they could be paid by Zway while a fee is charged
from the users. This would also overall enhance the usefulness of the saved posts feature
on Instagram while benefitting the content creators and the users.

It would be imperative to obtain consent and credit the relevant researchers whose work
towards recognising best practices for facilitation and decision making are utilised for
training the chatbot. The implied approach would be to filter through biased data before
using for training however, complete elimination of bias is still a challenge. Hence, it would
be crucial to inform the users about the potential biases in the output. The part where
users are chatting in group chat with AI, they could be sharing their personal thoughts,
preferences, and sensitive information. Again, a consent for allowing AI to store, analyse
information and provide suggestions based on that is a must (GDPR). Setting guardrails for
AI to limit the analyses and facilitation to travel planning related conversation would be
essential to maintain users’ privacy and as a respect for human dignity. Additionally,
transparency about the process and letting users own their data would be essential to gain
users’ trust. Though these are aspects which can only be tested and refined after extensive
training, they are noteworthy to be thought of. On top of the ensuring the user privacy and
security, a legal compliance would be another must while building this application.

Zway being an AI driven application would use users’ data for providing hyper personalised
and relevant recommendations hence it would be essential to prioritise user consent,
transparency, and legal compliance to ensure ethical handling of users’ data.
All in all, the insights from the research emphasise the need for a structured approach in
AI-assisted collaborative trip planning and the importance of aligning design practices with
emerging user preferences and technological advancements.
63

7. Conclusion
In conclusion, integration of AI and human intelligence can lead to superior outcomes
compared to what each could achieve individually in terms of collaborative group trip
planning. AI’s significant positive impact in the realm of remote work collaboration through
automating tasks, aiding in business decision making, tracking progress, and simplifying
search foreshadows its potential in group trip planning. Particularly, AI can assist with
group decision-making, search result personalisation, itinerary automation, and route
optimisation. The continued ethical advancement and improvement of AI technology point
to a promising future for the practical and safe use of the concept discussed in this study.

Currently, most applications focus on individual trip planning. However, a transition
towards applications for collaborative trip planning is anticipated given rapid advances in AI
and emerging needs. It is established that one size doesn’t not work for all even with AI
products. To address the unique needs of the defined users in this study, travel-related AIchat models must be improved and customised beyond the general search models now in
use. Users in this case looked for ease and efficiency in planning with hyper personalised
results and assistance in asynchronous collaboration. Therefore, AI's use in group trip
planning was well-positioned to satisfy these needs, providing innovative solutions for
these users that value personalised experiences and convenient collaboration. Since,
personalisation includes utilising users’ data, hence accounting for ethical measures
becomes a paramount consideration.

Furthermore, the nuances of designing with AI and gaps in process are identified through
this study. This equips designers to better understand the points of user’s interaction with
AI and anticipate the flows. Though the user flow and the idea can be tested with design
64

prototypes however, most of the user experience which involves AI interaction can only be
tested with ML enabled prototype.

Therefore, this study not only contributes to our understanding of AI’s capabilities in the
realm of travel but also serves as a roadmap with actionable insights for designers to
navigate a user centric AI-driven experience.

65

References
Bassett, B. (2023, September 23). The Peak-End Rule: Crafting Memorable User
Experiences. Nudge Notes. https://medium.com/nudge-notes/the-peak-end-rulecrafting-memorable-user-experiences-bf9a93de0056
Decrop, A. (2005). Group Processes in Vacation Decision-Making. Journal of Travel & Tourism
Marketing, 18, 23–36. https://doi.org/10.1300/J073v18n03_03
Fesenmaier, D. R., & Jeng, J.-M. (2000). ASSESSING STRUCTURE IN THE PLEASURE TRIP
PLANNING PROCESS. Tourism Analysis, 5(1), 13–27.
Framework for Innovation—Design Council. (n.d.). Retrieved February 12, 2024, from
https://www.designcouncil.org.uk/our-resources/framework-for-innovation/
Gen Z in travel: How to cater to this generation - Oaky. (2023, April 25). Oaky EN.
https://oaky.com/en/blog/gen-z-travel
Gonzalez, K. A. (2024). The Gen Z Travel Blueprint: What Gen Z Really Wants When Trip
Planning. encora. chromeextension://efaidnbmnnnibpcajpcglclefindmkaj/https://www.encora.com/hubfs/1.%
20Corp%20Thought%20Leadership%20Campaigns/Travel%2C%20Transportation%2
0and%20Logistics/GEN%20Z%20Research_5MB.pdf
Harding, X. (2022, February 11). When Structural Racism Feeds the Algorithm. Mozilla
Foundation. https://foundation.mozilla.org/en/blog/breaking-bias-how-algorithmicracism-goes-unaddressed/

66

Heaven, D. (2020, July 17). Predictive policing algorithms are racist. They need to be dismantled.
| MIT Technology Review.
https://www.technologyreview.com/2020/07/17/1005396/predictive-policingalgorithms-racist-dismantled-machine-learning-bias-criminal-justice/
Hwang, Y.-H. (2011). A Theory of Unplanned Travel Decisions: Implications for Modeling Onthe-Go Travelers. J. of IT & Tourism, 12, 283–296.
https://doi.org/10.3727/109830511X12978702284516
Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk.
Econometrica, 47(2), 263–291. https://doi.org/10.2307/1914185
Kim, S., Eun, J., Oh, C., Suh, B., & Lee, J. (2020). Bot in the Bunch: Facilitating Group Chat
Discussion by Improving Efficiency and Participation with a Chatbot.
https://www.researchgate.net/publication/339438103_Bot_in_the_Bunch_Facilitating
_Group_Chat_Discussion_by_Improving_Efficiency_and_Participation_with_a_Chatbot
Kore, A. (2020). Designing human-centric AI experiences: Applied UX Design for Artificial
Intelligence. Essay, A. https://learning.oreilly.com/library/view/designing-humancentric-ai/9781484280881/html/Cover.xhtml
Mason, M. C., Zamparo, G., Marini, A., & Ameen, N. (2022). Glued to your phone?
Generation Z’s smartphone addiction and online compulsive buying. Computers in
Human Behavior, 136, 107404. https://doi.org/10.1016/j.chb.2022.107404

67

Mucci, T. (2024, January 17). Chatbot examples: A beginner’s guide. IBM Blog.
https://www.ibm.com/blog/chatbot-examples-a-beginnersguide/www.ibm.com/blog/chatbot-examples-a-beginners-guide
Nichols, W. (2022). A Uses and Gratifications Approach to Travel Information Utilization: A
Generational and Gender Analysis. https://hdl.handle.net/2346/96524
Nielsen, N. (1993). Iterative Design of User Interfaces. Nielsen Norman Group.
https://www.nngroup.com/articles/iterative-design/
Ozdemir-Guzel, S., & Bas, Y. N. (2021). Gen Z Tourists and Smart Devices. In N. Stylos, R.
Rahimi, B. Okumus, & S. Williams (Eds.), Generation Z Marketing and Management in
Tourism and Hospitality: The Future of the Industry (pp. 141–165). Springer
International Publishing. https://doi.org/10.1007/978-3-030-70695-1_6
Park, K., & Roehl, W. (2016). Exploring Unplanned/Impulsive Travel Decision Making. Travel
and Tourism Research Association: Advancing Tourism Research Globally.
https://scholarworks.umass.edu/ttra/2010/Oral/40
Roeschke, L. (2023). Gen Z Travel Trends Could Reshape the Industry. Morning Consult Pro.
https://pro.morningconsult.com/analysis/gen-z-travel-trends
The gap between traveler expectations and industry perceptions (TRAVELER VALUE INDEX).
(2023). Expedia Group.
The impact of AI on the travel industry | McKinsey. (2023).
https://www.mckinsey.com/industries/travel-logistics-and-infrastructure/ourinsights/what-ai-means-for-travel-now-and-in-the-future

68

The Path to Purchase: Uncovering how travelers plan and book online. (2023).
https://go2.advertising.expedia.com/path-to-purchase-2023
Thi Loan, Ph. D. N., & Viet Hoang, Master. N. (2023). Generation Z’s Travel Trend and
Behavior: A Systematic Literature Review. JOURNAL OF ECONOMICS, FINANCE AND
MANAGEMENT STUDIES, 06(04). https://doi.org/10.47191/jefms/v6-i4-13
Young, I. (2008). Mental Models: Aligning Design Strategy with Human Behavior. Roseneld
Media, LLC.
Zhang, L., & Sun, X. (2016). Designing a Trip Planner Application for Groups: Exploring Group
Tourists? Trip Planning Requirements. 1329–1336.
https://doi.org/10.1145/2851581.2892301

69

Appendices
Comparative Analysis

Attributes/Apps Troupe
Value Proposition

Mitravel

Wanderlog

Tripit

PLAN GROUP Reimagine

Build, organize, and

TripIt helps you keep

TRIPS

map your itineraries

track of your flights,

group travel

WITHOUT THE planning!

hotels, car rentals,

HEADACHE.

confirmation numbers,
and other details–all in
one place–so you do
not have to search
through multiple
emails or apps to find
what you need.

Est Date

2020

Ratings on App Store 26 Ratings

2021

2019

2006

Not launched 500K+ downloads &

76.7K(Googleplay),

on App store

243,881(Apple)

9.99K ratingGoogleplay
1.3K rating- Apple
store.

Key Users
Tone & Copy

Any travel

Travel

groups

enthusiast

Clear and Crisp Fun &

Travel Planner

Frequent Travellers
Friendly & convincing

Friendly

70

User Reviews

“The entire

Not available "This is a fantastic app "Helpful in many cases

family is

for planning complex

but may get confusing

onboard, it’s

journeys in a wide

if you change

one central

region. It was easy to

bookings"

place to put

add points of interest,

"Two basic functions

everything and

and seeing them all on done well; the rest of

get out of

a map was immensely the app is

group text

helpful. I loved being

unimpressive"

chains…”-

able to group them

"I travel often both for

Derek G.

into itinerary days. And business and with my

Columbus, OH,

having all the flight

big family. This has

“When I sent

and hotel details in

been a game changer."

this app to

one place was super

people, they

useful! Worth pro to

felt more

take it offline for the

confident and

duration of the

onboard with

trip.""Updated, back to

the trip

2 stars. The core is

because

great, I like the

Troupe was so

integration with

organized.”-

Google maps and

Dondré T.

routing. But there are

Los Angeles,

annoying things like it

CA , “We’ve

will not route to/from

gotten more

airports/hotels even

done in the last

though they get

2-3 days with

marked on the map.

Troupe than

The design looks nice

we have since

in pictures but it's

March trying to

unclear in use, there's

plan this on our

no clear delineation

own.”-

between sections it's

Daphnee M.

all shades of grey."

Reading, PA
Good Features

Inviting friends Planning

Itinerary, Map view,

Live updates and

to plan

board, Polling downloadable

notification about

together. Poll

feature, AI

events

& voting for

Assistant for

decision

bookings

making.

71

Explore with AI
along with a
map. Feedback
on the
experience.
Poor Features

Add the links

Board view

and

on a mobile

Poor UI and navigation Doesn't allow group
planning for free

suggestions
manually. No
time limit on
voting feature.
Pricing

Free

Free-1 trip, 2

$5.33/month or

editors

$63.99/year

$50/year

$25/trip-5
editors
$12/month
Design

Strengths

Easy to use.

Poor UI and navigation Edit option in Organe

Too many

colour completely

clicks to

different from CTA.

navigate and

Lacks essential

plan

functions like duplicate

Heavily

Collaborative collaborative itinerary

Great App to remind

invested in

Planning on

flight and reserved

simplifying

board

options.

Group Travel
Weakness

Explore feature disorganised

overwhelming and too Not collaborative

using AI and

and difficult

many features

planning with

to track, No

groups are

flow

fragmented
experiences

72