You can create a hash that behaves like what you want. Hash.new takes a block which is invoked whenever a key lookup fails. We can create an empty hash when that happens:

hash = Hash.new { |hash, key| hash[key] = Hash.new(&hash.default_proc) }

hash[:first][:second] = [1, 2, 3, 4]

hash # => {:first=>{:second=>[1, 2, 3, 4]}}

Note though that merely accessing an inexistent key will result in the creation of a new hash:

hash.dig(:a, :b, :c) # => {}

hash # => {:first=>{:second=>[1, 2, 3, 4]}, :a=>{:b=>{:c=>{}}}}

hash[:foo].nil? # => false

I've assumed that the answer to the question I raised in a comment on the question is "yes".

One could use Enumerable#reduce (a.k.a. inject):

def undig(*keys, value)
  keys[0..-2].reverse_each.reduce (keys.last=>value) { |h,key| { key=>h } }   
end

undig(:first, :second, [1,2,3,4])
  #=> {:first=>{:second=>[1, 2, 3, 4]}} 

or recursion:

def undig(*keys, value)
  keys.empty? ? value : { keys.first=>undig(*keys.drop(1), value) }
end

undig(:first, :second, [1,2,3,4])
  #=> {:first=>{:second=>[1, 2, 3, 4]}} 

dig cannot be used to assign a value to a Hash, this method has been built only for accessing a value.

For your case, you can do one of both things :

hash = { first: { second: [1, 2, 3, 4] } }

Or :

hash[:first] = { second: [1, 2, 3, 4] }

You can also use the approach in that post : How to set dynamically value of nested key in Ruby hash

They create a new hash method to dynamically assign nested values to a Hash.