Generative AI, an ally for UX researchers.

Since the early days of user-centred design, user research has been an indispensable tool for creating digital products that truly meet people’s real needs.

As technology has advanced, useful tools for researchers have emerged, not only to speed up the process, but also to improve the quality of the research. We are now at a very important breakthrough: the integration of artificial intelligence throughout the research process.

According to the User Interviews report on the State of User Research 2023, 20% of researchers are already using AI in their processes and 38% plan to do so in the near future. This data shows that AI is becoming an increasingly relevant tool for user research.

In this post we look at how generative AI is redefining the dynamics of user research and paving the way to new insights.

How artificial intelligence is driving user research forward

Despite what some may think, generative AI should not be seen as a threat to researchers. Rather than taking away their work, it will help them do it more efficiently and more creatively.

Specifically, incorporating AI into the research process brings the following benefits:

1. It gives us a starting point. Generative AI can be used as an initial source of inspiration. Based on the prompts given by the researcher, it can provide a first structured basis on which to start. We are referring to tasks such as providing context on a topic, suggestions for scripts, interviews, questionnaires, etc.
2. Automate the most manual and repetitive tasks. In user research there are operational tasks that are absolutely necessary but do not add value in themselves. These are mechanical tasks such as writing emails to schedule interviews or survey instructions, taking notes during interviews, etc. Generative AI can help with reporting and hypothesis generation. This not only saves valuable time, but also allows professionals to focus on interpreting results and applying their knowledge.
3. Enhanced analytical skills. Generative AI can help us not only to process data quickly and accurately, thereby enhancing the analytical skills of researchers, but also to provide deeper and more revealing insights, further reducing researcher bias and enabling a more complete understanding of users’ needs and expectations.

In short, generative AI helps us speed up the process and reach deeper conclusions when dealing with large volumes of data.

But for now, while we know that generative AI has great potential for research, it is still an emerging technology and should be used with caution.

As a general recommendation, I would say that we can use it to support our traditional methods (in the following section we give you some ideas), but not so much to draw certain conclusions, as its results are not entirely accurate at the moment and this is where the experience and judgement of the researcher comes into play.

On the other hand, we should be very careful about using sensitive data with generative AI. Free consultations are used to train models, so avoid including names, addresses, ID numbers, etc. It is better to use the enterprise versions that the big players make available through their platforms, such as Google PaLM2 (instead of Gemini) or Azure Open AI (instead of ChatGPT) which guarantee confidentiality.

Where AI can add the most value in our research process (at the moment)

Based on the assumption that AI results are not entirely accurate in generating conclusions and that the researcher is key, we share some ideas where generative AI can be useful throughout the research process.

Ideas for the research plan approach

Once the researcher has gathered the necessary information and defined the research objectives, he or she can use generative AI to provide an initial structure of what the research plan will look like. AI therefore allows us not to start from a “blank canvas” and provides us with a good starting point with the appropriate methods and techniques for the research we want to carry out. But, as we have said, the researcher will then have to revise and adapt the plan according to his or her criteria and experience. For this initial task, generative AI tools such as Bard and ChatGPT can be very useful.

Help with desk research

For example, Gemini and ChatGPT (generative AI) can be very useful for certain desk research tasks. By providing them with the text of a document, article, etc., they can provide a summary and highlight the most relevant points for the study you are carrying out, or provide a context and explanation as an introduction to the topic you are researching, saving you a lot of time.

Qualitative research

For qualitative methods, generative AI can currently help us most in generating ideas for scripts, questions, interviews, transcription, text analysis and the generation of possible hypotheses. The last of these needs to be checked very carefully.

Interviews and focus groups

If, for example, we are planning to conduct interviews or focus groups, generative AI can also be of great help at various stages of the process:

• Screening. Based on certain instructions (the objective of the study, the characteristics of the profiles needed, specific topics, etc.) generative AI can give us ideas on what to ask in the screening questionnaire in order to better identify the profiles to be interviewed or invited to the focus groups. It can even help us draft the email that people receive before the questionnaire. Again, it is not necessary to start from scratch, but it will need to be reviewed and adapted.
• Write a basis for scripts and questions. Again, by providing the appropriate instructions (profiles to be used, objective, topics to be covered and explored in depth, time, etc.) generative AI can help us to design and structure an initial script.
• Capture tone, style and summarise conversations. Several tools have been available for some time that allow the researcher to avoid taking notes of conversations in interviews, moderated tests or focus groups. But now there are also tools that use generative AI to capture the tone and style of the conversation, summarise it and extract the key points, making their job much easier.
• Analyse the texts. Generative AI can help us generate labels, identify patterns and even analyse sentiment from transcripts to get categorisations, behavioural patterns and identify users’ motivations, whether they are excited or frustrated about a part of the process. There are tools that specialise in these tasks, such as Dovetail, tldv.io or Atlas.ti.

Usability test (qualitative part)

• Generate tasks. AI can help researchers to generate a set of tasks to be tested, which may be those that a user is most likely to perform and others that, although feasible, may be more challenging, outside what we consider to be the “Happy path” and undoubtedly interesting to test.
• Analyse the transcripts (the qualitative part of the tests). Here I would apply the same guidelines as above.

Quantitative research

When we need to obtain quantitative data in our research, for example to validate hypotheses generated in qualitative research or to identify patterns of user behaviour, AI also plays a role. In short, it can help us generate ideas for survey questions, analyse results, predict behaviour or even personalise experiences.

In general, it can be used to automate and enhance quantitative research tasks and improve the creativity of the actions to be taken, such as asking more innovative questions, identifying trends that are not apparent in more traditional analysis, making behavioural predictions that go a step further, or making personalised recommendations about content or experiences.

Surveys

If we consider the creation of surveys as part of our study, AI can help us in the following ways:

• Question generation. In the question generation phase, tools such as Jasper ai, ChatGPT or Gemini can again provide us with a first base of possible questions to include in the surveys, based on information provided by the researcher. For example, if the company uses tools such as Hotjar on its website, it would be interesting to use the ability to conduct surveys through this tool, as it will generate more relevant questions using contextual information from users.
• Analyse the results. When it comes to analysing survey results, the same products that use AI to generate surveys also use it to analyse user sentiment, satisfaction or trends, which are very difficult to detect with the naked eye when dealing with large volumes of data. But they can also go further and identify, for example, which types of customers are more likely to buy certain products, or even suggest product recommendations.

Usability test (quantitative part)

• Data analysis In the more quantitative part of usability testing. AI can help us perform automated data analysis to identify behavioural patterns, trends and key metrics in an automated way.
• Predicting usability problems. For example, it can identify patterns that suggest certain users may struggle with certain aspects of the interface.

Some AI tools that researchers can use

Although we have mentioned a number of tools throughout the post, Gemini and ChatGPT are the best-known and most useful generative natural language AI tools that researchers can use as a supplement to perform tasks such as:

• Generating ideas for research questions.
• Extracting insights from interviews and meetings.
• Assisting in the analysis of qualitative data.
• Writing texts.

However, these tools are not sufficient for all research tasks. For more specific questions, there are a number of specialised user research products that already incorporate AI. Here are just a few:

• Calendly: this tool helps us to organise interviews in the calendar, for example by analysing availability and suggesting meetings.
• Dovetail: a very useful tool for transcribing e.g. interviews, where you can tag and cluster relevant content, and even edit and trim videos of interviews to highlight a particular part of the interview in a report, for example. There are also popular transcription tools such as LoopPanel, Otter and Grain.
• Userdoc: can help us to create user stories and user journey proposals, among others.
• VisualEyes: enables eyetracking in testing processes.
• Hotjar: Hotjar is a web analytics tool that uses AI to provide information about user behaviour on a website. For research, it can be used to conduct surveys, for example, as it generates more relevant questions with contextual information from users.
• UserZoom: it is a tool that has been in use for many years and as part of its evolution now uses AI to automate data collection, pattern recognition and creation of visualisations after usability tests or surveys.
• Atlas.ti: is a super-complete tool for researchers, offering a complete set of AI-based functionality for analysing research data.

Finally, we encourage researchers to use generative AI in conjunction with and as a complement to traditional methods. However, it is essential that practitioners use their experience and judgement to review and adjust any analysis or recommendations. This technology is advancing day by day, and as researchers we need to remain vigilant, testing and exploring these methods as they undoubtedly have and will have a long way to go.