A year ago we posted a thought piece around best practices in data management and reporting to increase trust and value from our data. These were summarised as the Seven Cs of Compliance, Confidence, Consolidation, Consistency, Clarity, Context and Causation and we hope they are still pertinent to your understanding and use of data.

A year is a long time in data. Since last August we have seen the ‘rise of AI’ and we can hardly open a newspaper (if you still do ‘open a newspaper’) without seeing headlines about threats and opportunities of AI.

With this in mind SiQ have revisited the Seven Cs and considered how these same best practices can help shape and steer our understanding and use of AI tools. We hope this is of interest – please get in touch to share your thoughts:

• Compliance, in adhering to standards to ensure data integrity, is crucial when handling data used in AI. Compliance through industry standards as well as in-house rules helps avoid biases in AI algorithms and ensures transparency in data processing and AI outputs. By embedding compliance rules, we can foster trust and mitigate potential risks associated with AI data use and misuse.

• Confidence in data used in AI is essential for reliable and accurate outcomes. We need to be sure of the quality, completeness and relevance of the data that we input. Rigorous data validation processes, including data cleaning, normalization and verification, enhance confidence in the input data. Additionally, establishing data governance frameworks and monitoring of data sources and collection methods contribute to confidence in the ongoing use of that data. By instilling confidence in AI data, we can make informed decisions, improve predictive accuracy and build trust in the AI systems we develop and deploy. Without this confidence we should be very wary of the outputs we receive.

• Consolidation, we commonly work with varied data sources. Data from multiple sources, such as structured and unstructured data, used to input AI should be brought together into a unified dataset. Data consolidation enhances the comprehensiveness and representativeness of the AI models, enabling more accurate predictions and insights. It also facilitates data exploration and discovery, uncovering hidden patterns and relationships. By consolidating varied data, we can make more informed decisions based on a comprehensive understanding of the data available to us.

• Consistency of data used in AI is crucial for reliable and valid outcomes. Consistent data ensures that AI models are trained on dependable and accurate information, reducing the risk of biased or misleading results. Achieving consistency involves data standardization, normalization, and regular data quality checks. By embedding and maintaining data consistency, we can improve the performance and trustworthiness of AI systems, enabling more robust decision-making and enhancing the overall effectiveness of AI applications.

• Clarity of queries used in AI is crucial for generating accurate and relevant outputs. We need to be sure to provide clear and unambiguous questions or commands to extract meaningful information from AI systems. Well-defined queries enable AI models to better understand our intent and context accurately, improving the precision and validity of the generated outputs. Clear queries also help mitigate potential misunderstandings or misinterpretations, ensuring effective communication between users and AI systems.

• Context provides the necessary background information and situational awareness for accurate interpretation and comprehension of AI-generated outputs. AI models heavily rely on context to deliver relevant and meaningful responses that better align with our intent, so excluding context or getting context wrong can severely deviate any AI process. Establishing context in our AI use fosters better user experiences, reduces misunderstandings and enhances the overall effectiveness of AI applications in meeting our requirements.

• Causation poses a significant challenge for users in interpreting AI responses. AI models excel at pattern recognition and correlation, but causation is a more complex concept for AI tools to grasp. AI can identify relationships between variables but may struggle to determine the underlying cause-and-effect relationships. Interpreting AI responses without a clear understanding of causation can lead to costly, incorrect assumptions and unreliable insights. It is crucial to approach AI outputs with caution, acknowledging AI limitations and the need for human intervention to properly interpret causation. Combining AI applications with expertise from experienced teams and trusted partners can help bridge the gap to enable more accurate interpretation of and better use of AI applications.

About Scholarly iQ
Celebrating 20 years, SiQ provides unrivalled expertise for trusted usage analytics. SiQ’s fast, easy, platform independent implementation delivers valued data insights to customer, sales, editorial, marketing and partner stakeholders. Questions about COUNTER, Open Access, Impact, Data or wider Business Intelligence? Trust SiQ to have you covered.