The ECOVAL framework provides guidelines and the GAMBIT tool to increase ecological validity in usability testing. ECOVAL defines attributes like objects, tasks, and environment to objectively measure ecological validity. GAMBIT allows producing low-cost, realistic prototypes for remote usability testing. A case study showed GAMBIT improved ecological validity over paper testing, enabling earlier usability problem detection and increased productivity. The authors recommend using ECOVAL and GAMBIT to improve usability evaluations and realize cost savings and other benefits from more ecologically valid testing.
Diamond Application Development Crafting Solutions with Precision
ECOVAL: A Framework for Increasing the Ecological Validity in Usability Testing
1. Suzanne Kieffer
ECOVAL: A Framework for Increasing the
Ecological Validity in Usability Testing
Jean VanderdoncktUgo Braga Sangiorgi
Université catholique de Louvain
Louvain School of Management
Louvain-la-Neuve, Belgium
4. Internal validity
Was the study well done?
External validity
Can we generalize the findings to
particular persons, settings and times?
Ecological validity
To which extent can we generalize the
findings?
7. What’s the problem with ecological validity?
No formal definition that enables
To objectively quantify the ecological validity of
experimental designs
To capture the critical attributes of the real-world
environment
9. Outline
The ECOVAL framework
ECOVAL
Guidelines
GAMBIT
Case study
Operability and usefulness of ECOVAL
ROI of increased ecological validity
Recommendations
24. • Q1: To which extent is ECOVAL
applicable for controlling the
ecological validity?
• Q2: What are the possible
outcomes/benefits of
increased ecological validity ?
28. Monitoring report
Task Control Result
Task Control Result
Artefact:
paper checklist
Section > Subsection
Function – Name - Date – Shift
Section > Subsection
300 items a day
29. Monitoring report
Task Control Result
Task Control Result
Artefact:
paper checklist
Section > Subsection
Function – Name - Date – Shift
Section > Subsection
42. Paper GAMBIT
Ecological
validity
1 1.17
Data
collection
Manual & face-to-face
Difficult & tedious
Unstructured process
Obtrusive
Automated & remote
Comfortable
Structured process
Unobtrusive
Outcomes
Compliant with
the literature
Early detection of usability problems
Increased user productivity
Decreased late design changes
Increased organizational efficiency
Good morning everyone! My name is Suzanne Kieffer, I’m from UCL, BELGIUM, and I’m very pleased to attend my first HICSS conference.
During the next 20 minutes I will present ECOVAL, a framework built for increasing the ecological validity of experimental designs.
In particular, I will focus on the applicability of the concept of increased ecological validity in usability testing.
First of all, what is ecological validity?
In experimental research, the term ecological validity refers to the correspondences between an experimental design and its related natural setting.
Especially in terms of users’ environment, stimuli and tasks.
The ecological validity of a study is high if it takes place in the participants’ natural setting or if the participants don’t know they are taking part in a research study.
On the contrary, it is low if the experimental design differs from the situations, stimuli and tasks experienced in a natural setting.
How does ecological validity relate to other types of validity?
The overall validity of an experiment can be broken down into three types of validity: INTERNAL, EXTERNAL and ECOLOGICAL validity.
And the overall validity can be compared to a puzzle. If one piece is missing or has been damaged, that will negatively impact the final result.
For example, a study may have been adequately carried out, with a HIGH INTERNAL VALIDITY (instrumentation, selection, experimenter).
However, if the experimental design did not sample the variables from the user’s environment, the study may suffer from problems with ecological validity.
And consequently, the generalizability of the findings may be reduced.
And this is exactly the reason why ecological validity is so important to anyone intending to generalize the conclusions from the particular circumstances of an experiment to wider ecological situations.
Why is ecological validity important in HCI as well?
BECAUSE it will drive user interface design
By impacting the OVERALL validity of experimental studies
And therefore by impacting the conclusions, outcomes and findings of usability studies
As more or less ecologically valid
That’s why we propose ECOVAL
A framework for defining, controlling, increasing the ecological validity in usability testing.
Which brings me to the outline of me presentation
The ECOVAL framework builds on the combination of Schmukler’s three dimensions of ecological validity and Shackel’s components in human-machine systems.
The signals and objects dimensions both refer to the stimuli used and describe how the real world environment is captured in the experimental design.
Signals : sounds, smells, colors, lighting, heat, cold, dust, and pressure
Objects : buildings, equipment, furniture, tools, and artifacts
Test medium and user interface dimensions both refer to the computer system being tested.
Task and behavior dimensions both refer to what is required of the participants.
Similarly to wireframes, Gambit enables the EASY, QUICK and COST-EFFECTIVE production of prototypes
With the advantage of performing as a real product user interface in terms of INTERACTION, NAVIGATION and RESPONSIVENESS
GAMBIT displays the screens of the user interface, enables the navigation between them, and it can run on multiple devices: smartphone, tablet, laptop, monitor, TV, etc.
The first step is the production of screens.
Here they were created with Balsamiq, but they could have been created by simply talking a picture or scanning a sketch, or even by producing screenshots from a higher-fidelity prototype
The second step is the creation of the behavior of the prototype, by simply interlinking specific areas of images
The third step is to conduct the experiment
Again, prototypes can run on multiple devices
And the last step is the download of the log file and the analysis of data
Framework + companion guidelines + supporting tool
So the research questions are:
We conducted a case study in a company whose core-business is hot-dip galvanizing to produce steel sheet for the automobile industry
What’s important to know about Hot-Dip galvanizing is that the production line runs continuously through a cleaner, an annealing furnace, and then into a molten zinc bath.
And that it requires the collaboration of 7 co-workers: one team leader, one quality technician, five operators.
USER TASK = COMPLETION of the MONITORING REPORT
FILL-IN a checklist on a daily basis
300 items a day
USER TASK = COMPLETION of the MONITORING REPORT
FILL-IN a checklist on a daily basis
300 items a day
We identified 5 key users: the team leader, and four operators.
And this is the timeline we followed:
We conducted user profile, task analysis, usability goal setting, and screen design
What’s important here is that
that the same user interface was evaluated in both the paper and the GAMBIT evaluations
And that the ecological validity was higher with the GAMBIT prototype than with the paper prototype with a ratio…
Problem new version of the prototype
What we believe enabled to identify the usability problems and correct them, is the fact that the GAMBIT prototype was more realistic/ecological in terms of test medium and task.
The intended device
The task
The use of mock objects lead to the user walking from one mock object to another
I will now present the most important results and discuss them
First of all,
Increased ECOVAL enabled the early detection of usability problems
By doing so, it enabled to correct these usability problems
Which led to increased user efficiencies and increased user satisfaction
Secondly, it led to savings in terms of both usability engineers and users productivity
To determine the costs we need to know the number of hours spent with the GAMBIT experiment
The increased productivity was calculated as the operator-hours saved per year (375 hours) multiplied by the operator wage (80). The decreased late design changes were calculated as the costs of late changes minus the costs of early changes.
Yet, there are a number of other questions beyond the scope of this paper that need to be addressed in near future.
What correspondences can be made between behaviors in an experiment and behaviors in the real environment?
Or how to transfer the natural/intended behaviors in an experimental context?
A top-down ecological approach to tackle that relies on synthetic signals [1, 17]. Real signals are first recorded in the field. Then, participants’ behavior, efficiency and satisfaction towards the synthetic versions of these signals are studied in the lab. Relevant signals in the case study are alarms, phone calls, and GA line stop.
How do signals and behavior dimensions exactly impact external validity?
Finally, this raises further questions. Is there a threshold in ecological validity that, once reached, ensures external validity?