This document discusses formulating better hypotheses for abductive design research. It explains that hypotheses should be testable, have independent and dependent variables, and state a clear relationship between them based on evidence. Poor hypotheses can lead teams in the wrong direction, so it's important to go back to the scientific method. The document provides templates for generating hypotheses for generative and evaluative research. It also outlines steps for formulating, testing, and assessing hypotheses as part of an iterative research process. Formulating the right hypotheses is presented as key to effective abductive design.
20240507 QFM013 Machine Intelligence Reading List April 2024.pdf
Formulate better hypotheses
1. Formulate better hypotheses
For a more effective Abductive Design
Carmen Brion | @Tea_monster
UX Oxford and DOPM | 10-05-2016
2. Abductive design takes the best guess
Abductive Reasoning
● Starts with an incomplete set of
observations
● Forms hypotheses making a
prediction based on the information
you have at the moment
● Goes from an observation to a theory
that frames the most likely
explanation for the observation
Contrary to popular opinion - Sherlock Holmes rarely, if ever,
deduces anything. He may occasionally induce something,
but most of the time he infers the best explanation from his
observations. Source
Picture by Graham
3. Hypotheses make design research
collaborative and targeted
Facilitates teams to:
● Discuss and align about the
problems to solve
● Separate assumptions from facts
Picture by James Stencilowsky
6. This is not a hypothesis, it is just a
description of how pagination works.
If online shoppers use pagination then they
will have more items to explore
If online shoppers use pagination then they
the likelihood of buying increase
If online shoppers use pagination then they
will have different pages of items to explore
This is not a hypothesis, states a fact
and it is not worth investigating
This is a hypotheses but uses a solution
as independent variable and it’s too
vague
7. Poor hypotheses are worse that no hypotheses
Like poor research, give teams a false sense
of going in the right direction.
Bad hypotheses lead to
● Designing the wrong experiments
● Ambiguous data coming out of the
research
● Results are difficult to interpret and
translate into outcomes
Picture by Quinn Dombrowski
8. To learn to formulate
good hypotheses
We need to go back to
the scientific method
Picture by Lefteris Heretakis
9. How to make
hypotheses sound
The good hypotheses:
1. Are simple and unambiguous
2. Have a dependent and an
independent variables
3. State a clear, unique relationship
between these 2 variables, based on
some evidence/guess
4. Are testable, measurable, falsifiable
and positive
Anyone reading the hypotheses should
be able to understand the context and
clearly interpret the results.
Picture by Gerry Lauzon
10. Pagination is a solution and this
limits the outcomes of the research
Diverge
Converge
Having the independent variable as the solution makes
us converge too early and limits what we can find out
in the research.
Converge
Diverge
Having the independent variable solution-free makes the
research divergent. The results will help to understand the whys
behind this relationship and define the solution that fits better.
If online shoppers use pagination then their likelihood
of buying increases
If we give online shoppers a bigger amount of attractive
items then their likelihood of buying will increase
11. Understand where you are in
product development journey
Who is the audience? What are their problems/needs? How can we solve it? What’s the best solution?
Who are the personas?
Are there potential new
personas?
What are the Job To Be Done?
What are their key behaviours?
What are the personas problems?
What are their pains and
frustrations?
How can we solve these
problems?
What are the barriers and/or
constraints?
How different solutions solve
the problems?
How the different solutions fit
with the personas mental
model?
Which solution is better?
How is the chosen solution
working?
What is the impact to the
personas?
Generative Research
Evaluative Research
“Find the root cause, don’t
just put a plaster over the
symptoms”
12. Generative research
hypotheses templates
Audience Problems Solutions
We believe that [this persona(s)]
are a potential new audience
because of [this reason]
We believe that [this persona(s)]
will behave this way because of
[this reason]
We believe that [this persona(s)]
with these goals have [this job
to be done]
We believe that [this persona(s)]
have [this problem] [achieving this
goal] because of [this reason]
We believe that [this persona(s)]
have [this pain/frustration]
[achieving this goal] leading to this
[consequence]
We believe that [this problem] will
be solve [with this solution] that will
result in [this outcome]
We believe that [this persona(s)]
using [this solution] will result in
[this outcome]
We believe that [this solution] will
result in [this change in the
persona(s) mental
model/behaviour]
“Ask the right questions to
get insights that lead to the
right solutions”
13. Evaluative research hypotheses templates
Solution
If we provide [this solution] then [this persona
(s)] will be able to achieve [this outcome] and
results in [this improvement]
If [this solution] is the best solution for [this
persona(s)] to achieve [this outcome] as a
results we will see [this improvement]
If-then hypotheses are more suited for
evaluative research. However, we can
improve them to make sure they meets the
good hypotheses requirements.
14. Hypotheses will evolve as we run research and
we learn more about the audience needs
We believe that fashion shoppers need more items in a category
because they are looking for inspiration
We believe that fashion shoppers find it hard navigating a
great amount of content
We believe that we can help fashion shoppers finding items
they want to buy by relating their search to their lifestyle
If the Fashion category has lifestyle articles then
the likelihood of fashion shoppers to add 1 or
more items to the basket will increase a 20%
16. Steps to formulate, test and assess hypotheses
Determine the type of
questions to answer
Do we need evaluative
or generative research?
Determine questions
What do we know, don’t
know and what are the
problems/constraints
Determine
assumptions
Formulate hypotheses
Following the
guidelines
Prioritise the
hypotheses
Which ones are more
risky if they were
wrong?
Design the research
To assess the
prioritised hypotheses
Run the research
Assess hypotheses
Validate/invalidate
hypotheses with the
results of the research
Capture the
learnings
1 32 4 5 6 7 8 9
17. Use Test Cards from the
Value Proposition Design
● State assumption
● Outline the task to test the assumption
● Think how you are going to measure
success
● Indicate what success means
Will increase the likelihood of fashion shoppers
finding items of their interest.
Having a lifestyle filter
will be asked to find items With a regular filter
and with the lifestyle filter
Participants will search
A 20% more of items added to the basket
than regular filters
Lifestyle filters should achieve
Number of items they add to the basket
18. Use Learning Cards from the
Value Proposition Design
● State assumption
● Explain what was observed for this
assumption during the research sessions
● Extract the learning
● Indicate the action to take in line with the
learning
Will increase the likelihood of fashion shoppers
finding items of their interest.
Having a lifestyle filter
that are more attractive but fashion shopper
might not add them to the basket
because they want to try it first.
Lifestyle filters generate Items
the non-linear user journey to get fashion shopper
back for items they had shown interest on.
The user journey is not linear.
Lifestyle filters
Review how to address
19. For listening :)
“Few ideas work on the first try.
Iteration is key to innovation”
― Sebastian Thrun
21. If you use hypotheses, please let me know
how it goes.
Please feedback:
● What works and how it has impacted your work
● Any tips or cool stuff you would like to share when putting hypotheses into practice
● What is hard to put into practice and what problems you are encountering
I am planning to keep working to make this work better, so I am planning to keep iterating these ideas
and practices. Your ideas and feedback are always welcome, the more brains the better problem
solving!