Artificial Intelligence Starts with Trusting Your Data

Artificial Intelligence (AI) and the technologies associated with it have disrupted the market and created room for bigger and better opportunities. Based on the insights from the data, these technologies require a lot of trust and assurances. Since organizations are expected to trust the authenticity of these insights, the data needs to be clean and in line with the expectations that are associated with it.

All stakeholders associated with the organization, including employees and management, should understand the value of keeping data clean, regardless of whether it is in an unstructured or structured form.

SAS recently conducted a business survey with over 2,400 business leaders from across the globe. The survey was conducted to delve deep into the details pertaining to best practices for fair AI implementation and trusting the data organizations already have.

Being a proud member of the SAS Collaborator Programme, I was granted access to the results of this survey. The results were a variety of perspectives all leading to different sides of the same coin.

While most of the business leaders surveyed as part of this study wanted increased access to data, only a few were able to say with confidence that they had the right data with them to make the decisions that their organizations need. This just builds on the dearth of accurate data sources in the market, with many organizations and business leaders having to work without accurate data sources.

Leaders that showed a high level of interest in using data for analytics and generating insights were also more likely to use a wider variety of methods to ensure that the data they had met the quality standards they wanted. These quality standards are important to the cause, as they can help generate useful insights.

In short, it is important to have access to the right sources of data to access or create the kind of trust you need to make important decision for your organization.

Why Trust is Important and the Role of Innovation

As companies start adapting AI and associated fields, their understanding of the systems, including deep neural networks, becomes critical to the cause. However, we may not always be able to identify why a certain machine learning or AI model might have taken a specific action. Since it is human nature not to trust what the human mind doesn’t understand, we tend to distrust AI and its potential to bring solutions to our problems.

This distrust can only be negated through the presence of trust and confidence in the data collection process and the assimilation method for the AI model. Organizations should be able to trust the data they are putting into the system and leverage their trust in that every data to build an even deeper connection.

Innovation is crucial here, as it will eventually open up the AI black box that currently hinders organizations from trusting some models. The need of the hour is to create an AI system that can perform reliably every day; an innovative AI system that gives consistent results on consistent data entry. The results should be replicable and reliable. The AI system also needs to be transparent, so it is possible to understand how it makes decisions so imperfections – or even bias – can be removed.

The Trust Gap

While there are multiple reasons as to why people in organizations currently feel like they don’t have the correct data to work with, the survey by SAS probed further into the matter and found evidence pointing to a certain trust gap in the market.

The gap is that only a very small percentage of users surveyed as part of the research mentioned that they could always trust their data based on the qualities of completeness, relevance, accuracy and timeliness.

These results give organizations a great chance to speed up their data quality assurance measures and build the confidence required for flawless data. The results do indicate that organizations can easily do more and enjoy greater laurels from the market. Only around 20 percent of all respondents surveyed as part of the research mentioned that they performed routine monitoring and formal approaches to maintain data quality and consistency.

However, experts mention that rather than fixing data after it has been collected, organizations should turn their attention to ensuring quality while collecting data. Jeanne Ross, a research scientist at MIT believes that the worst time to fix your data is after it has been collected. Your full focus should be on improving the business processes that lead to the data coming in.

Ross further elaborated that, “while it’s straightforward to say, fix your processes so that the data collection is reliable, and the quality issues are pretty minimal meeting that goal is a challenge. That’s because it takes ongoing discipline to refine data collection processes, testing data quality regularly along the way.”

The trust gap can only be filled by running data quality measures at home. Organizations have realized the need for stringent checks on the data streams coming in and have not only added to the frequency of these streams but are also looking to ensure seamless data quality across the organization.

How to Build Trust in AI?

Explainable AI is the key need to bridge the gap between analytics and their perceived benefits. An explainable AI strategy will make more sense in reducing bias and building your trust in the system. Organizations and scientists will generally have better trust in an AI system that they can build and tweak, as opposed to a black box that you will never be able to change.

Analysis augmented by explainable AI will help advance the outcomes of the decision-making process by giving more food for thought.

Conclusions

The research by SAS reached the following conclusions for future growth:

Better Data Governance Is Desperately Needed

Only a small minority of respondents from the survey had initiated formal activities for data quality assurance. This minority points towards the imperative need for data governance for better and advanced analytics.

The practitioners interviewed as part of the process were quick to provide examples for improving data structure, quality, governance and architecture to build trust in raw data.

Data Privacy Is an Opportunity

Data privacy was one of the leading concerns for survey respondents. They believed that there were enough opportunities present in this regard to improve the security practices pertaining to the application of analytics. That current data privacy initiatives aren’t as strong and capable as they should be was a key highlight from the interviews. Respondents were also keen to further use GDPR and other privacy initiatives to raise trust levels within their customer base.

Fostering an Analytics Culture Can Improve Innovation

By creating a culture that is driven by analytics organizations can improve their innovation. This was mentioned by numerous respondents, alongside the need for a skilled workforce. Both metrics were considered imperative to increasing organizational innovation.

You can download the Data, Analytics & AI: How Trust Delivers Value report to discover more about the findings by SAS as part of this survey.