Lower scientist shooting accuracy?
ai.npc_use_new_aim_system "false"

Assembly-CSharp.dll --> BaseProjectile:

    public override Vector3 ModifyAIAim(Vector3 eulerInput, float swayModifier = 1f)
    {
        float single = Time.time * (this.aimSwaySpeed * 1f + this.aiAimSwayOffset);
        float single1 = Mathf.Sin(Time.time * 2f);
        float single2 = (single1 < 0f ? 1f - Mathf.Clamp(Mathf.Abs(single1) / 1f, 0f, 1f) : 1f);
        float single3 = (0 != 0 ? 0.6f : 1f);
        float single4 = (this.aimSway * 1f + this.aiAimSwayOffset) * single3 * single2 * swayModifier;
        ref float singlePointer = ref eulerInput.y;
        singlePointer = singlePointer + (Mathf.PerlinNoise(single, single) - 0.5f) * single4 * Time.deltaTime;
        ref float singlePointer1 = ref eulerInput.x;
        singlePointer1 = singlePointer1 + (Mathf.PerlinNoise(single + 0.1f, single + 0.2f) - 0.5f) * single4 * Time.deltaTime;
        return eulerInput;
    }​

 

I don't know if you're making a mod, or just trying to lower scientist accuracy:

ai.npc_valid_aim_cone  defines how close their aim needs to be on target in order to fire. (default: 0.8) (0.8)

ai.npc_valid_mounted_aim_cone   defines how close their aim needs to be on target in order to fire while mounted. (default: 0.92) (0.92)

ai.npc_deliberate_hit_randomizer The percentage away from a maximum miss the randomizer is allowed to travel when shooting to deliberately hit the target (we don't want perfect hits with every shot). (default: 0.85f) (0.85)

ai.npc_deliberate_miss_to_hit_alignment_time The time it takes for the NPC to deliberately miss to the time the NPC tries to hit its target. (default: 1.5) (1.5)

ai.npc_alertness_to_aim_modifier This is multiplied with the current alertness (0-10) to decide how long it will take for the NPC to deliberately miss again. (default: 0.33) (0.5)

I went in console and entered find aim  find cone  find accuracy  find ai to find these.