Building your own product is high-risk, high-reward. But, if it works, you can benefit from a positive feedback loop that enables distinctive research.
Taking the approach of building your own experiment one step further, some researchers actually build their own products. These products attract participants, and then serve as platforms for experiments and other kinds of research. For example, a group of researchers at the University of Minnesota created MovieLens, which provides free, non-commercial personalized movie recommendations. MovieLens has operated continuously since 1997, and during this time 250,000 registered users have provided more than 20 million ratings about more than 30,000 movies (Harper and Konstan 2015). MovieLens has used the active community of users to conduct wonderful research ranging from testing social science theories about contributions to public goods (Beenen et al. 2004; Cosley et al. 2005; Chen et al. 2010; Ren et al. 2012) to addressing algorithmic challenges in recommendation systems (Rashid et al. 2002; Drenner et al. 2006; Harper, Sen, and Frankowski 2007; Ekstrand et al. 2015); for a full review see Harper and Konstan (2015). Many of these experiments would not have been possible without researchers having complete control over a real working product.
Unfortunately, building your own product is incredibly difficult, and you should think of it like creating a start-up company: high-risk, high-reward. If it is successful, this approach offers much of the control that comes from building your own experiment with the realism and participants that come from working in existing systems. Further, this approach is potentially able to create a positive feedback loop where more research leads to a better product which leads to more users which leads to more researchers and so on (Figure 4.15). In other words, once a positive feedback loop kicks in, research should get easier and easier. Despite the potential upside to this approach, I can’t find any other examples of success, which shows just how difficult it is to execute successfully. But, my hope is that this strategy will become more practical as technology improves. These difficulties with creating your own product mean that researchers who want to control a product are much more likely to partner with a company, the topic I’ll address next.