Explore insights with an AI-powered assistant

Conjointly’s Insights Explorer is a utility for processing and analysing data using the R programming language directly in your web browser. It functions as an integrated development environment software (IDE) for rwasm, a third-party module of binary R packages for WebAssembly. With pre-written R scripts called recipes, you can run analyses and visualisations with a few clicks, without writing any code and installation.

The Insights Explorer is available for free, and does not require a Conjointly account to access. Conjointly users can access the Insights Explorer through the Utilities menu or directly from their launched experiment reports. Use and distribution of the rwasm software is subject to external licences.

Through the large language model (LLM) analysis interface you can analyse data and create visualisations by describing what you want to know in everyday language. The LLM will generate the appropriate code which you can run to produce desired insights and visualisations without coding.

You can refine your analysis through follow-up prompts, and export any visualisations for external use. Each analysis also comes with interpretation guidance and suggestions for deeper exploration of your data.

Open Insights Explorer

Example prompts

Simply enter your analysis request and the chat interface will generate R code to run it. Here are some prompt examples and what they do:

Getting started

Access the LLM chat interface on the Chat tab in the top-left corner.

This interface works with your experiment data, which you can load in two ways:

• Load experiment data via the Files tab.
• Launch Insights Explorer directly from your experiment reports for automatic data loading.

Making analysis requests

Write your analysis requests in the chat input field. Being specific in your request helps get better results. For instance, specify which variables to compare or what type of visualisation you prefer.

The LLM interface will display an initial system message which contains two main components:

1. Default system prompt
   This is the default prompt that instructs the LLM how to process your requests. For example:
   

   You are an AI agent that has access to R. You can execute R code directly on the user’s device. You have been given access to user’s experiment on the Conjointly market research platform in R format. User needs you to perform data analysis. They will give you requests. You will helpfully write R code (with concise explanation). You will only write one chunk of R code per response inside the ``` notation.

2. Summary of current data environment
   This shows what data is loaded from your experiment and available for analysis:

   • All loaded R objects in your current session
   • Available data tables from your experiment
   • Column names for each table

Based on your request, the system provides:

1. A brief explanation of what it will do
2. Generated R code for the intended analysis, including:
   • Required library imports
   • Data preparation steps
   • Analysis or visualisation code
3. Detailed explanation, including:
   • What the code does
   • How to interpret the results or visualisations

The following shows an example response when requesting a correlation analysis.

Request: Plot a correlation matrix using number of AI tools, AI concerns, purchase intention, gender, and age

Response:

# Load required library
library(ggplot2)

# Select the specified columns
selected_columns <- c(
  "Q5: _ AI tools usage",
  "Q25: _ AI Concerns",
  "Q30: _ Purchase intention",
  "Q36: Gender",
  "Q37: Age (write-in)"
)

# Create a subset of the data with selected columns
subset_data <- rawDataList$Respondents[, selected_columns, with = FALSE]

# Function to convert columns to numeric
convert_to_numeric <- function(x) {
  if(is.factor(x) || is.character(x)) {
    return(as.numeric(factor(x)))
  } else if(is.numeric(x)) {
    return(x)
  } else {
    return(as.numeric(x))
  }
}

# Apply the conversion function to all columns
subset_data <- as.data.frame(lapply(subset_data, convert_to_numeric))

# Calculate the correlation matrix
cor_matrix <- cor(subset_data, use = "pairwise.complete.obs")

# Create a data frame for ggplot
plot_data <- data.frame(
  x = rep(colnames(cor_matrix), each = ncol(cor_matrix)),
  y = rep(colnames(cor_matrix), times = ncol(cor_matrix)),
  correlation = as.vector(cor_matrix)
)
# Create the correlation plot
p <- ggplot(data = plot_data, aes(x = x, y = y, fill = correlation)) +
  geom_tile() +
  scale_fill_gradient2(low = "#F8696B", mid = "white", high = "#63BE7B", 
                       midpoint = 0, limit = c(-1,1), space = "Lab", 
                       name="Correlation") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust = 1)) +
  coord_fixed() +
  geom_text(aes(label = round(correlation, 2)), color = "black", size = 3) +
  labs(x = "", y = "", title = "Correlation Matrix of Selected Variables")

# Display the plot
print(p)

Please review the response to ensure it meets your analysis needs. If modifications are needed, simply type your feedback or additional requests in the chat input field. The system will generate a new response incorporating your feedback.

Executing the analysis

When you are ready to run your analysis, click on >_ Run and circle back .

In the Code from assistant popup window, please review the R code snippet and click on the >_ Run and circle back again to execute the generated code. The system will run the analysis and display your results.

If any errors occur, don’t worry. The LLM will suggest fixes in its next response, which you can then review and execute the analysis.

Viewing results

After executing the code, the system:

• Generates the requested visualisations or analysis
• Provides a summary of key findings
• Highlights notable patterns or trends in your data
• Recommends further actions or analysis

You can review these insights and make follow-up requests for deeper analysis if needed.

The following example shows how the LLM presents the visualisation, results, and recommendations after executing the correlation analysis code above:

Based on the correlation matrix and plot, we can interpret the relationships between the variables as follows:

• AI tools usage (Q5) correlations:
  • Moderate positive correlation with Age (0.26)
  • Weak negative correlation with AI Concerns (-0.13)
  • Very weak positive correlations with Purchase intention (0.05) and Gender (0.05)
• AI Concerns (Q25) correlations:
  • Very weak positive correlations with AI tools usage (0.05), Age (0.06), and Purchase intention (0.04)
  • Very weak negative correlation with Gender (-0.03)
  • Purchase intention (Q30) correlations:
  • Weak positive correlation with Gender (0.15)
  • Weak negative correlations with Age (-0.12) and AI tools usage (-0.13)
• Gender (Q36) correlations:
  • Weak positive correlation with Purchase intention (0.15)
  • Weak negative correlation with Age (-0.11)
  • Age (Q37) correlations:
  • Moderate positive correlation with AI tools usage (0.26)
  • Weak negative correlations with Purchase intention (-0.12) and Gender (-0.11)

These insights can help in understanding the relationships between different aspects of AI perception and usage among the survey respondents. However, it’s important to note that correlation does not imply causation, and further analysis might be needed to draw more definitive conclusions.

Key insights:

1. The strongest correlation is between AI tools usage and Age (0.26), suggesting that older respondents tend to use AI tools more frequently.
2. There’s a weak negative correlation between AI tools usage and AI Concerns (-0.13), indicating that those who use AI tools more tend to have slightly fewer concerns about AI.
3. Purchase intention has a weak positive correlation with Gender (0.15), suggesting a slight tendency for one gender to have higher purchase intentions for AI-related products or services.
4. Age has weak negative correlations with both Purchase intention (-0.12) and Gender (-0.11), implying that younger respondents might have slightly higher purchase intentions and that age distribution might differ slightly between genders.
5. Overall, most correlations are weak, indicating that there aren’t strong linear relationships between these variables.

Overall, the correlations between these variables are generally weak, with the strongest correlation being -0.17 between gender and purchase intention. This suggests that while there are some relationships between these variables, they are not strongly predictive of each other.

Note that these analyses and interpretations are generated by an LLM. Always verify results before making decisions based on the analysis.

You can click the button in the top-left corner of the chat interface to select New File to open the R code editor for custom analysis, or New Chat to start a new conversation.

Next steps

You can also read about:

• Analysing your data using Recipes
• Importing data into Insights Explorer
• Writing and running R code in Insights Explorer
• Viewing variables in Insights Explorer