Each of the wires will generate its own magnetic field and you could calculate those magnetic fields individually, and use the right-hand rule to work out in which direction the magnetic fields circulate (not vice-versa), and then add up the magnetic fields (as vector fields) to work out the net magnetic field at any point.
Ampere's law tells you that you don't necessarily have to do that if you just want the line integral of the B-field. What matters then is just the net current passing through a closed loop. The direction of the magnetic field along that loop will depend on the net current interior to the loop and on the geometry of the loop and how close it passes to the individual wires. Thus in your picure, I would guess that if $I_1 + I_3 > I_2$ then the B-field is likely to circulate counter clockwise (looking down from the top).
A caveat here is that this is not a very symmetric current distribution. The influence of $I_2$ will be greater closer to the wire carrying that current and vice-versa. Ampere's law only tells you the total line integral, not the value or the direction of the field at any point, except in highly symmetric situations.