How to analyse PromptAI Reponses

PromptAI is a question type in SurveyMR 2.0, which automatically prompts a participant’s initial verbatim answer for more detail in a natural and engaging way, using generative AI technology. PromptAI responses can be analysed using the TextMR tool.

Note: A minimum of 50 responses is required per data set in order to run coding analysis within TextMR.

How do I access TextMR?

To access TextMR, go to the Back Room and navigate to Analysis Tools > TextMR using the left-hand menu.

Creating a TextMR analysis

To create a new TextMR analysis, click on ‘Add Text Analysis’ on the top right of the page.

Selecting your source

  1. Add a Title for your analysis. You can also add an optional
  2. Then, select a Source Type. The Source Type is the tool type that was used to collect the data. In this case, as PromptAI is a question type in SurveyMR 2.0, the Source Type will be SurveyMR 2.0.
  3. Then select a Source. The Source is the name of the task that was used to collect the data. In this case, the Source will be the name of a survey.

 

4. Click ‘Save’.

5. You will then see a new page. On this new page, select the survey question you’d like to analyse, using the drop-down field, then click ‘Add question’.

6. (Optional) To Add multiple sources, you can then click ‘Add more sources’ in the top right of the page. This will allow you to add additional questions from your selected source, which can later be sent for coding.

Coding your responses

Once you have selected your source and selected ‘Add question’, the ‘Responses’ tab will be generated automatically and all responses for the selected question will be listed. As highlighted in the yellow box below, you will also see the number of responses, and how many are coded / uncoded. In the example below, all responses are currently uncoded.

At this stage, within the Responses tab, you have the option to exclude any specific verbatims within TextMR if you don't want them to be coded (e.g., they contain PII).

Responses which contain PII will automatically be highlighted by the TextMR tool. In the example below, this is shown by the phrase [REDACTED_EMAIL) in place of an email address which was given as part of this particular answer.

  1. You can then code these responses by clicking the ‘Code responses’ button.

2.   A box will then appear, to ask if you would like to proceed, with a reminder that this will use ‘X’ amount of AI Credits

Each response coded will use one AI credit only, including responses which are gathered via a PromptAI SurveyMR 2.0 question, as both the user’s initial response, and their additional response after being prompted, are coded as one.

 

3.  To proceed, click ‘OK’. The page will then show that responses are being coded, and you should check back later

For surveys with a large number of responses it may take a few moments for these to appear, and for very large datasets, this may take up to 3 hours.

Accessing your analysis

Once your responses have been coded, in addition to ‘Summary’ and ‘Dashboard’ tabs, you will gain 3 forms of analysis within your Responses tab which are listed below:

Theme

The theme of a response is a compressed version of the users’ response and begins the process of grouping responses together based on common themes.

Netcode

Netcode further categorises your responses into broad categories based upon the way in which the user has responded.

Sentiment

Positive, negative and neutral sentiment tags are added to signify the general sentiment of the users’ responses.

In the Responses tab, you can filter based on any of these codes within the search bar shown below:

 

Summary tab

Once your responses are coded, you will gain a Summary tab which will neatly summarise the common themes and feelings from the cumulation of responses to your survey question. The length and detail of the summary will differ based upon the number of responses and the level or differentiation in user responses. As an example, the below is a screenshot of a summary of 50 responses to the question ‘How do you feel about the size and pricing of chocolate bars now compared to ten years ago?’

Sources

Once your responses have been coded, you can return to the Sources tab within TextMR to add additional data sources to your analysis set (in this case, additional PromptAI questions, from either the same survey, or a different survey). You can then send any new source(s) for coding within this analysis set.

*When adding new data to an analysis that has already coded responses, any new data is coded using the existing codes. If you wish to generate and analyse by new codes, you will need to create a new analysis in order to do this (just make sure that you add all the sources that you want to analyse to your analysis before you select to ‘Code responses’, as described above).

Dashboard

The Dashboard tab presents the coded data in the form of tables, allowing you to visualise coded responses as quantifiable data. These tables consist of: Themes, Netcodes, and Sentiments.

Themes

As well as compiling the common themes from coded responses, this table also shows you the percentage of coded responses that contained this theme.

 

Sentiments

Positive, negative and neutral sentiments are displayed in a simple and clear pie chart to give you the general sentiments of netcodes at a glance.

 

Netcodes

A simplified view of netcodes is displayed in a bar chart and gives you information on the categories that themes are grouped within, as well as the percentage of coded responses within each netcode.

 

 

Downloading your analysis

Whilst TextMR is designed to analyse your results in the InsightHub, we understand that you may want to download these results and use these externally. To generate a download of your analysis you will first need to return to the list of analyses on the main TextMR page. Then, select the dropdown option under the Actions column, within which you will find the ‘Download XLSX’ option which will generate your download.

The XLSX-type download will contain columns with the following headers:

  • Source type (Tool type, for example, SurveyMR 2.0)
  • Source (Name of activity)
  • Question
  • User answer ID
  • Response
  • Theme
  • Netcode
  • Sentiment
  • User ID
  • Screenname
  • Columns for any InsightHub user fields which are ticked to ‘show in CSVs'

Deleting an Analysis

On the main TextMR tool screen, where you can see a list of TextMR analyses, each TextMR analysis has a drop-down button in the ‘Actions’ column. This is where you will find the option to delete an analysis.

*An alert will appear when attempting to delete data linked to an ActivateMR board, in case you require the data to remain.