An Agenda for Disinformation Research

Speaker(s): Garland, J. (Santa Fe Institute)

Date: 20 July 2021

Speaker Session Summary

SMA hosted a speaker session with Dr. Joshua Garland (Applied Complexity Fellow, Santa Fe Institute) as part of its SMA IIJO Speaker Series.

Dr. Garland defined disinformation as “false misleading or inaccurate information disseminated with the intent to deceive.” Disinformation must be countered because it undermines individuals’ trust in US institutions and destroys society’s shared reality, increasing polarization and societal conflict. Furthermore, technological innovations and a lack of media literacy have allowed disinformation to increase its impact and reach. Dr. Garland identified six important topics for future research to counter disinformation: a) detection of disinformation at scale, b) measuring the impact of disinformation, c) open research data infrastructure, d) new ethical guidelines, e) educational interventions, and f) workforce training.

Disinformation is difficult to detect at a large scale because of the information environment’s massive ecosystem and constantly evolving technology. Also, fact checking is currently not something that artificial intelligence (AI) can do well, which makes countering the massive amount of disinformation more difficult and reliant on individuals. Effective disinformation exploits AI’s weaknesses, such as sarcasm, irony, or subtlety. There is no current way to measure the impact on disinformation on either the individual or societal level, because there are too many independent variables. Dr. Garland mentioned the need to create an open research database for researchers to share information on disinformation. The current list of ethical guidelines for behavioral research comes from the 1978 Belmont Report, which does not relate well to the new digital sphere. He proposed two solutions for countering disinformation: educational interventions throughout society and training computer professionals to consider the potential misuse of their algorithms.