Infection Simulation based on Proposed Human Behavior Models derived from SNS and Press Data

Research and Development by

Satoshi KURIHARA, Faculty of Science and Technology ,Keio University

Corresponding Research Area

Simulation and designing countermeasures against possible COVID-19 resurgence: predicting spreading of infection, estimating and verifying the effectiveness of countermeasures, and predicting deman

Simulation using Multi-Agent Simulation etc. :SNS Analysis

Simulation using Multi-Agent Simulation etc. :SNS Analysis

■The appearance of keywords from TV reports and SNS (Twitter) which are thought to promote human behavior change are analyzed, and the human behavior model is estimated.


②Behavior Model Estimation

Consider the transformation of human behavior by dividing it into transformation 1 (anxiety and sense of crisis) and transformation 2 (socio-economic behavioral desires).

Transformation 1 occurred in the 1st wave (Anxiety Coefficient) → Vague sense of crisis about unknown and not-well-understood COVID-19 → Several infection examples (and the celebrities’ obituary) became a big topic. State of Emergency triggered Transformation 1 → Wearing a mask and avoiding 3Cs are promoted (Infection Control Coefficient is increased).

Transformation 2 is occurring after the 2nd wave→ Because the infection mechanism and the detailed status are clarified, the sense of crisis disappears and Transformation 1 no longer occurs.→ Transformation 2 is socio-economic behavioral desire (Action Coefficient) → Traveling for work (minimum economic activity) does not stop, whereas traveling for sightseeing is highly refrained → Trade-off between “the restrictions of masks and avoidance of 3Cs” vs. “free movement”. 

③Behavior Model Simulation

How to promote self-restraint?
  • Increase Anxiety Coef. →Systems directly conveying the risk of infection to individuals (i.e. COCOA) is effective (α).Mass media is likely to be ineffective unless they are exaggerated considerably. Short-term repetition of enacting and cancelling self-restraint (rather than pinpoint self-restraint).
  • Increase Infection Control Coef. →Wearing a mask and avoiding 3Cs. There might be some other options
  • Decrease Action Coef. (is difficult) →Consideration is required for policies encouraging traveling (i.e. GO TO TRAVEL campaign) (β).

Infection simulation of traveling between urban and local city based on a human behavioral change model

Case study

Starting the simulation with an outbreak in Tokyo and Osaka and no cases in Naha and the outlying islands.
※Naha and outlying islands cause outbreaks due to high-risk behavior of movement of infected people from Tokyo and Osaka

Even infrequent travel, accompanying high-risk behaviors, leading to the spread of infection and pressures on health care in non-infected areas.


Increase Infection Control Coef. →Wearing a mask and avoiding 3Cs.
Increase α → Policy encouraging self-restraint Policy
Increase Anxiety Coef. → Diffusion of COCOA (e.g., institutionalization of installation)

TODO: Validating the effects of the self-restraint policy

(1) Repeating whole self-restraint in period of ’t’ and release in period of ‘m'.
(2) Continuing self-restraint longer in some urban only.