Back Share
Strategies

Questions to consider in key stages of AI and machine learning based research, regarding methodological scope

project-design reflection-questions

“Assessing Alternative Methodologies and Scope of Research Internet experimentation projects can be scaled to a worldwide level (e.g. Kramer, Guillory & Hancock, 2014) and engineers are typically incentivized to deploy a project as widely as possible to maximize their reach. It is sometimes also just easier to let a project operate without limitations and to see later which data is collected, rather than limiting its scope artificially. However, the knowledge gained using this collection method can have exposed some problems in specific political and cultural contexts. Risk levels can vary widely based on target countries or for particular target groups (see also Dwork, 2006). Therefore, trying to mitigate the risks and shifts in values in all areas will result in a race to appease the lowest common denominator (or: reduce the utility of the project to appease the context with the highest risk factors). - How can the researcher limit the scope of the research questions and the project’s aim to avoid some risks of harm or negatively affected values? - How can the researcher limit the scope of stakeholders, by excluding particular groups or countries? If so, would the data collected still be a representative sample to answer the research question? - Are any risks averted if the researcher limits the project duration to a shorter amount of operation time? And does this conflict with the ability of the researcher to conduct the research in question?” (franzke et al., 2020, p. 40)

Overarching Principles Merit and Integrity
Sources AoIR report 3
Title Questions to consider in key stages of AI and machine learning based research, regarding methodological scope