Skip to content
Interface language: English
Design / Signal

Two people, one task: when to study an interface in pairs

Design a paired usability task around real cooperation, information handoffs and a shared outcome rather than counting two independent successes.

Article language: English

Illustration of two people at a table before a large circular structure and mountains.
In this article

In a hypothetical test, an exhibition coordinator chooses a display case while a technician checks whether it fits by the entrance. She knows the arrangement of the exhibits; he knows the width of the available route. If we seated each person separately and gave them all the answers on a sheet, we could test use of the catalogue. We would not see how they pass along a condition on which their joint decision depends.

Studying a task in pairs makes sense when collaboration is part of use. It can reveal questions, hints and misunderstandings between two roles. Two people being present does not by itself guarantee better material. One may take over while the other confirms choices they do not independently understand.

Let us prepare a test for a fictional exhibition equipment planning tool. Ola coordinates the project and Janek prepares the room layout. They must agree on and save a display case choice. This is a proposed research scenario, with no test performed and no real participant results.

A pair does not mean two independent users

In a moderated usability test (in Polish) (in Polish), a facilitator observes a task being performed. Here we also observe how participants pass necessary information to each other. We do not add a second person to an ordinary test afterwards just to make speaking aloud feel more natural.

First establish the dependency. Ola knows which exhibits need a particular case. Janek has the room plan and knows the entrance constraint. Both need data available in the tool, but neither can make the entire decision solely from their own role. This division should correspond to real work, which must be understood beforehand.

A GOV.UK interview guide permits talking with pairs or small groups when people use a service together. This is guidance for understanding the context of shared use, not a ready-made test protocol. Before preparing the task, we can establish who actually needs information from whom.

The user, buyer and payer model (in Polish) (in Polish) describes roles in a service. For paired research we choose specific users who collaborate. The person paying for a tool need not participate in planning display cases. Business influence over a purchase does not automatically give them the appropriate role in this test.

Record whether the people know each other and how they usually work. A regular pair may have their own shorthand; people meeting for the session may need more explanation. Do not present a random combination of two volunteers as a reconstruction of real collaboration. You can deliberately study a first meeting, but that is a different question.

Also describe differences in experience. Janek may have known the catalogue for a long time while Ola is using the tool for the first time. That pair suits a question about collaboration between these roles if it reflects the work. Do not mix it without identification with a pair of two beginners. Smooth completion may come from one participant's memory. Notes need to show when the application supplied information and when a person brought it.

A report by Pawson and Greenberg on research at a trade fair describes situations in which one participant spoke with the facilitator while the other continued the task alone. This is an observation from that setting. For our project it is a reason to record actual participation instead of assuming every minute of a paired session was joint work.

Choose the physical conditions too. If Ola and Janek normally work on two devices, seating them before one screen changes their access to information. If they normally look at a plan together, separating them completely may create a difficulty absent from their work. Materials and workstation arrangement should follow the collaboration being studied.

Before the session, perform the task in both roles using prepared test accounts. Check their permissions and starting material. If Janek cannot open the plan because the prototype is misconfigured, that is not yet an observation of product difficulty. Decide when to stop the test to repair the environment and when to leave participants with a real problem. Record the boundary so the same technical fault is not interpreted differently across pairs.

Prepare a task where the handover can be seen

In our scenario, Ola chooses a case appropriate to the exhibits, Janek checks the space, and both must agree on a saved option. Give each person information corresponding to their role. Do not hand them a common sheet containing all the answers in advance if passing information is what the test is meant to study.

Establish the completion condition. A choice saved in Ola's account need not be visible to Janek. Pair success should mean both recognise the agreed option and know whether the decision has been saved. It is not enough for the coordinator to say "we have it" while the other person politely stays silent.

Give each role its own starting point, but do not invent constraints solely for an impressive test. If both normally see dimensions on the plan, do not take them away from Ola just to make her ask Janek. We are studying a handover arising from the work. In our example we separate knowledge of exhibits and the room while keeping case data available in the catalogue according to the designed permissions.

Do not order them to swap roles halfway through merely so each clicks equally often. If the division of work is real, an unequal number of clicks may be normal. If you want to check interchangeability, prepare a separate test and name that objective. A forced swap creates different conditions from an ordinary joint decision.

On one device, note who controls it and when control is handed over. The other person may point, read a passage or open their own material. Do not reduce their contribution to click counts. If one person begins doing everything for the other, retain that sequence. Follow-up questions after the task can establish whether this is a normal division of work or a workaround for an interface difficulty. Unequal participation alone does not determine the cause.

Aeneia conceptual photography

For remote work, reproduce the available ways to communicate. Conditions for remote participation in a hybrid meeting (in Polish) (in Polish) illustrate the importance of shared material and the opportunity to speak. In the task test, add the question of which screen each person sees and how they pass missing information. Being on a call alone does not provide the same view.

The facilitator should not become a messenger who is absent from ordinary work. If Janek asks what Ola chose, first observe how the pair can establish it. Supplying the answer would rescue the task but conceal the handover problem. If help is necessary, record when it was given and its scope.

With two screens, account for different states. Ola may save a new option while Janek still sees the old one. That need not be a conversation error. Observe how each recognises the choice's freshness and whether they can establish whose change applies. In the example, distinguish a view that has not refreshed from a case where Janek saw the current option but understood it differently. Those events need different fixes.

In co-design (in Polish) (in Polish), participants may propose a way to agree beforehand. Paired research tests its use. Do not mix the stages: explaining one's own idea at a board differs from using a prototype without the idea author's guidance.

At Aeneia Design, the test design can be connected to both roles' screens and shared-choice states. It needs a description of the real information split and a decision about what must be agreed. Only then can the prototype show handover points without replacing them with the facilitator's commentary.

Aeneia Design

Let us map a jointly performed task: each person's knowledge, the screens and the moment the result is agreed. Together we can prepare a test that reveals both interface difficulties and missing information between roles.

Discuss a test of a jointly performed task

Record the pair's journey, not just time to finish

A note should let someone reconstruct an event: Ola identified an option, Janek asked for dimensions, the screen showed another version, and an explanation followed. This is more useful than the label "participants became confused". A fix requires knowing whether the problem was the information, its visibility or mutual understanding.

A note template for our task can distinguish two cases. In the first, Ola selects case S while Janek still sees M. After refresh both identify S and correctly state its dimensions. This is a proposed example of a note pointing to view freshness, not an observation from a conducted session. In that case we need to check transmission of the change and the signal that it appeared.

In the second, both see S but Janek takes the whole case's dimension for its interior dimension. Refresh resolves nothing. We ask about naming and presentation of dimensions and how the condition is explained to the other person. Only after that distinction can appropriate fixes be prepared. "The pair needed help" would lead to one general instruction even though the described difficulties have different causes.

Separate help from the other person from help from the researcher. If Janek knows the catalogue and explains the name himself, the task may be completed through his knowledge. That is a valid outcome of shared use if this division of roles occurs in reality. It does not prove that a new coordinator could handle that name alone.

A study by Alhadreti and Mayhew of a library website found no significant task-performance differences between individual and paired methods, while the paired approach required more analysis time. This result is limited to that implementation. It does not justify promising that doubling the number of people at a screen will always improve research.

Define the unit when counting results. A pair agreeing on one case is one joint trial, not two independent successes. Ola's and Janek's individual answers can be retained separately, for example when asking about their understanding of the outcome. Do not double the denominator of completed tasks merely because the recording contains two voices.

After the task, ask each person for a short individual explanation of what was selected and what will happen next. A joint conversation can then help explain differences. Initial separate answers can reveal that matching nods did not mean the same decision.

In analysis, the same event can have two effects. Ola completed the choice, but Janek does not recognise its status; the pair has not yet achieved the agreed outcome. Retain both descriptions instead of averaging one person's success and the other's failure. A joint task needs a decision about the whole journey. Individual answers explain where it diverged, but do not replace that criterion.

If you need to assess independent task repetition after a break (in Polish) (in Polish), plan it separately. Janek's hint may help Ola now, but it does not prove she remembered the route on the next visit. A test of collaboration is not automatically a test of individual learning.

The report should say what needs improvement for both roles and where another kind of test is still needed. In our example, the final question is whether Ola and Janek can choose the same case without a third person explaining their screens. The acceptance condition is a shared, recognisable result. The report should show what the interface enabled and where extra explanation was still needed.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear