Don’t trust AI? Trust My Explanation

Keren Yaniv /

AI computer vision Product

Customer trust is any company’s lifeline. To achieve trust, AI products need to alleviate the natural mistrust we have for black boxes

One of the first steps we took in the early days of Constru was to define our core values. From day one, Trust was critical and something that was brought up by every member of the leadership team, each from their own unique point of view. The term trust also repeated in many customer discussions – ranging from “how can we trust your system” – to “we need an objective data source, as we don’t trust our current solution.”

Lack of trust between all parties involved is a major issue in the construction industry, and one of the main problems to be solved in order to graduate to the next phase. So we decided to explore trust a bit deeper, to see how we make it an inherent part of our product.

When referring to trust, there are two main types that we can talk about:

  1. Practical:
    • Meet your commitments
    • Act upon your words
    • Prove your competency
    • Get the job done
  2. Emotional:
    • People trust that you’re on their side
    • You won’t judge them for their setbacks
    • They can feel comfortable telling you their honest thoughts, feelings and ideas

When it comes to AI, suspicion is even greater. Most people don’t really understand how artificial intelligence works, with some not even believing it really exists. The truth is that AI is very much a reflection of the company developing it, and the quality and effectiveness is also very dependent on them. In fact, the EEU even developed ethics guidelines for trustworthy AI to help regulate this challenge.

Companies that utilize AI are working towards the goal of being trustworthy by becoming more transparent about what decisions the technology makes for their customers. One study even found that people who were given the ability to modify an algorithm, were more satisfied with the results - even when they didn’t make any changes. They also said they were more likely to use AI again.

Constru approaches trust issues in AI by giving clients a way to feel in control of the data they receive, and the decisions computers make for them. So in what areas does AI effect Constru’s product? Great question! And the answer is “everywhere”.

Our product provides a deep analysis of the reality capture coming from the jobsite. It relies on computer vision algorithms understanding what is seen in the images, and comparing that to the 3D architectural model, schedule and other project plans. The Constru platform then leverages this data to provide valuable insights and project metrics to enable informed decisions that keep projects on schedule and under budget. We also provide reports aggregating the data and highlighting important calculation results.

We are often asked “how did you get to this number,” so we decided to add to our progress tracking reports a component which shows the exact status over a floor plan, including a representative image explaining why this progress was detected. When progress states 98% completion, and the customer is questioning the number, they can easily find that one element which was not completed. With this, we are exposing the core of how we do things, giving customers full transparency into the results of the computer vision detection, as well as the project plans as we have them updated in our system.

Another great example is coming from our discrepancy list implementation. When we detect a discrepancy, we allow the customer to have a full record of its status before and after it was resolved, as well as have full tracing capabilities if it was decided not to fix it. This allows also having full visibility into decisions made in the field and their reasoning, providing another layer of transparency to the project’s work.

We keep looking for ways to increase trust with our system and within the project team. If you have more thoughts and suggestions of where to add it, we would be more than happy to hear it.