Metrics of productivity can be valuable in assisting evaluation, but to do this they must provide complete and accurate descriptions of citations (Cousijn et al. 2018). Currently this is not the case. Fixing the problem is deceptively simple for a variety of reasons, one being that there is no consensus on how to handle self-citation data. The ongoing debate is contentious and further complicated by the widespread use of the h-index for research evaluation, which as the dominating metric puts the emphasis squarely on citations to guide decision making (Hirsch 2005; Hicks et al. 2015). Without question, this creates a real career motivation to strategically use self-citation (Seeber et al. 2019), but this does not in any way diminish the value of self-cites that result from productive, sustained, leading-edge efforts (Cooke and Donaldson 2014). When used appropriately, self-cites are equally important as cites from the surrounding community, and without tracking them it is impossible to see how scholars build on their own work.
Despite this, many favor a curated form of the h-index as a response to the gaming problem. Curation involves hacking away at the citation data to neatly remove all occurrences of self-citation. While such treatment effectively silences direct attempts to boost citation scores, it does not prevent indirect manipulation and also produces undesired side effects. For example, curation ignores when authors use self-citation to attract cites from others, which is alarming given that each self-citation appears to increase the number of citations from others by about one after a year, and by about three after 5 years (Fowler and Aksnes 2007). Furthermore, curation unfairly punishes good citation practices, a particularly worrisome issue for those publishing novel ideas or results that challenge well-established dogma. In such cases, self-citation data can be critical as paper outputs may require a substantially longer period of time to attract the attention (i.e., citations) they ultimately deserve. Thus it is not good practice to hide self-citation data. The end result is a distorted record of progress and discovery.
The sensible alternative to curated scorekeeping would be to consider all citation data including self-citations. Towards this goal, we demonstrate an easy way to track self-cites without distorting other metrics, namely the h-index. The approach is not meant to criminalize self-referencing, nor do we intend to suggest a certain threshold of acceptable behavior like what Hirsch did when proposing the h-index (Hirsch 2005). Rather we see this as a tool to clarify how researchers build on their own ideas, as well as how self-citing contributes to the bibliometric impact of their own work. Furthermore, researchers are less likely to blatantly boost their own citation scores (Zhivotovsky and Krutovsky 2008; Bartneck and Kokkelmans 2011) while others are watching.